Pipeline inner wall state detection method and device based on multi-feature fusion

Abstract

The invention provides a multi-feature fusion-based pipeline inner wall state detection method and device, and relates to the technical field of pipeline inner wall state detection.The method specifically comprises the steps that a to-be-detected pipeline is divided into a plurality of sections, an inner wall sample image of each section is collected and preprocessed, the defect area is represented by the pixel number, and the defect area is calculated; calculating a first evaluation coefficient in combination with the wall thickness data; acquiring historical operation parameters of a pipeline made of the same material as the to-be-detected pipeline and corresponding residual service life of the pipeline, and generating a sample data set; and constructing a pipeline residual service life prediction model based on a deep learning algorithm, inputting geometric parameters and physical parameters of a to-be-detected pipeline into the model, obtaining a residual service life prediction value, performing correction by using a defect evaluation coefficient, obtaining a fatigue state evaluation index, comparing the fatigue state evaluation index with a judgment threshold, and evaluating the state of the inner wall of the pipeline. According to the method, the accuracy and the real-time performance of pipeline inner wall state detection are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline inner wall state detection, and specifically provides a detection method and device for pipeline inner wall state based on multi-feature fusion. Background Art

[0002] In the modern industrial field, pipeline systems, as important infrastructure for liquid and gas transportation, play a crucial role. However, over time, due to long-term exposure to multiple environmental influences such as fluids, pressure, and temperature changes, the inner walls of pipelines are prone to defects such as corrosion, wear, and cracks. These defects not only affect the normal operation of pipelines but may also trigger serious safety accidents such as leaks and explosions, posing great risks to enterprises and public safety. Traditional pipeline detection methods mainly rely on manual inspections and regular detections, which are not only time-consuming and laborious but also often unable to comprehensively and effectively identify potential problems on the inner walls of pipelines. Due to the lack of efficient on-line monitoring means, many pipelines may experience accidents before obvious problems occur, resulting in huge economic losses and damage to the reputation of enterprises.

[0003] Existing technologies have obvious deficiencies in pipeline state monitoring and evaluation. Firstly, the limitations of traditional detection methods make it difficult for technicians to obtain real-time pipeline operation status and defect information, often relying only on historical data for judgment, resulting in maintenance decisions lacking timeliness and accuracy. Secondly, existing methods usually fail to effectively integrate multiple influencing factors such as environmental parameters, geometric parameters, and material properties, leading to a lack of a comprehensive evaluation perspective. In addition, when dealing with and analyzing image data, existing technologies often lack advanced image processing algorithms and are unable to accurately identify and quantify minor defects on the inner walls of pipelines, thus affecting the accuracy of evaluation results. Generally speaking, these deficiencies limit the effectiveness and reliability of pipeline detection, and there is an urgent need for a new type of multi-dimensional pipeline inner wall state detection method to improve the scientific nature and effectiveness of pipeline management.

[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a detection method and device for pipeline inner wall state based on multi-feature fusion to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solutions:

[0007] A detection method for pipeline inner wall state based on multi-feature fusion, the specific steps include:

[0008] Step 1: Divide the pipeline to be detected into several pipeline sections of equal length. Collect sample images of the inner wall of each pipeline section. After preprocessing the sample images, use the number of pixels to represent the defect area in the sample images, and combine the wall thickness data of the pipeline section to calculate the first evaluation coefficient. Calculate the average value of the first evaluation coefficients of all pipeline sections as the defect evaluation coefficient of the pipeline to be detected. The wall thickness data includes the original wall thickness data and real-time wall thickness data of the pipeline;

[0009] Step 2: Obtain the historical operation parameters of several pipelines with the same material as the pipeline to be detected. At the same time, obtain the remaining service life of the pipeline corresponding to each set of historical operation parameters. Map the historical operation parameters and the corresponding remaining service life of the pipeline one by one to generate a sample data set. The historical operation parameters include operation time, geometric parameters, and environmental parameters. The geometric parameters are specifically the pipeline diameter, pipeline length, and pipeline smoothness. The environmental parameters are specifically the average temperature, average pressure, and fluid pH value;

[0010] Step 3: Build a prediction model for the remaining service life of the pipeline based on a deep learning algorithm. Use the historical operation parameters in the sample data set as input and the corresponding remaining service life of the pipeline as labels to train the prediction model for the remaining service life of the pipeline to obtain a trained prediction model for the remaining service life of the pipeline;

[0011] Step 4: Real-time collect the operation time, geometric parameters, and physical parameters of the pipeline to be detected and input them into the prediction model for the remaining service life of the pipeline. The model outputs the predicted value of the remaining service life of the pipeline to be detected. Use the defect evaluation coefficient to correct the predicted value of the remaining service life to obtain the fatigue state evaluation index of the pipeline to be detected;

[0012] Step 5: Set a judgment threshold for the fatigue state evaluation index. Compare the fatigue state evaluation index of the pipeline to be detected with the threshold. According to the comparison result, evaluate the inner wall state of the pipeline and guide the management decision-making.

[0013] Furthermore, the preprocessing of the sample images includes noise removal processing and contrast enhancement processing;

[0014] Among them, the specific logic for noise removal processing is as follows: Use the wavelet transform denoising method to denoise the sample images. Decompose the image after distortion correction through wavelet transform to obtain wavelet coefficients of the image at different scales and directions; Perform threshold processing on the wavelet coefficients, set the wavelet coefficients with low amplitudes to zero, and retain the wavelet coefficients with high amplitudes; Perform inverse transform on the wavelet coefficients after threshold processing to reconstruct the processed coefficients into an image to complete the image denoising processing;

[0015] The specific logic for contrast enhancement processing is as follows: find the minimum gray value L in the sample image to be processed max and the maximum gray value L max , and apply the following linear transformation formula to complete the contrast enhancement processing:

[0016]

[0017] In the formula, I new (x, y) is the pixel value at the coordinate (x, y) after contrast enhancement processing, and I old (x, y) refers to the original image pixel value at the coordinate (x, y), and L is the number of gray levels.

[0018] Furthermore, the specific logic for using the number of pixels to characterize the defect area in the sample image is as follows: set a threshold according to the defect characteristics and convert the gray image into a binary image. The expression is as follows:

[0019]

[0020] In the formula, I binary (x, y) represents the value of the generated binary image at the position (x, y). If the pixel belongs to the defect area, then I binary (x, y) is 1, otherwise I binary (x, y) is 0; I gray (x, y) represents the value of the original gray image at the position (x, y), and T is the defect threshold set according to the experiment, which is used to distinguish the defect area and the normal area in the image;

[0021] Use the connectivity algorithm to find all connected defect areas and mark them. For each marked connected area, calculate its area, which is obtained by counting the number of pixels with a value of 1 in the binary image:

[0022]

[0023] In the formula, A defect represents the number of pixels in the defect area, P(j) represents the number of pixels in the j-th defect area, j is the index of the defect area, and N is the total number of defect areas;

[0024] At the same time, obtain the spatial resolution of the image, that is, the actual area represented by each pixel, and convert the number of pixels in the defect area into the actual defect area through the following formula:

[0025] A actual = A defect * A pixel

[0026] In the formula, A actual is the actual defect area, Apixel is the spatial resolution of the image;

[0027] Calculate the defect ratio based on the actual defect area and the total area of the pipeline section. The formula is as follows:

[0028]

[0029] In the formula, A is the defect ratio, A actual is the actual defect area, A total is the total area of the pipeline section.

[0030] Furthermore, obtain the wall thickness data of the pipeline section. Combine the defect ratio of the corresponding pipeline section. After dimensionless processing of the wall thickness data and the defect ratio, calculate the first evaluation coefficient. The formula is as follows:

[0031]

[0032] In the formula, AS is the first evaluation coefficient, A is the defect ratio, Thi0 is the original wall thickness data of the pipeline, Thi is the real-time wall thickness data, k1 and k2 are preset weights, k1>k2>0, and k1 + k2 = 1;

[0033] The wall thickness data includes the original wall thickness data and the real-time wall thickness data of the pipeline;

[0034] Calculate the average value of the first evaluation coefficients of all pipeline sections as the defect evaluation coefficient of the pipeline to be detected, denoted as

[0035] Furthermore, the specific logic for constructing the prediction model of the remaining service life of the pipeline is as follows:

[0036] Obtain the historical operation parameters of several pipelines with the same material as the pipeline to be detected. At the same time, obtain the remaining service life of the pipeline corresponding to each set of historical operation parameters. Map the historical operation parameters to the corresponding remaining service life of the pipeline one by one to generate a sample data set. Randomly divide the sample data set into a training set and a test set. The historical operation parameters include operation time, geometric parameters, and environmental parameters. The geometric parameters are specifically the pipeline diameter, pipeline length, and pipeline smoothness. The environmental parameters are specifically the average temperature, average pressure, and fluid pH value. Build a prediction model for the remaining service life of the pipeline based on a deep learning algorithm. Use the historical operation parameters in the training set as input and the corresponding remaining service life of the pipeline as labels to train the model to obtain a trained prediction model for the remaining service life of the pipeline. Substitute the historical operation parameters in the test set into the trained model to obtain the corresponding prediction results. Calculate the error between the prediction results and the actual values in the test set. Determine whether the error meets a preset error threshold. If it meets, output the trained model, that is, the prediction model for the remaining service life of the pipeline. If it does not meet, return and continue training. The error is the mean absolute error, root mean square error, and coefficient of determination between the prediction results and the actual values in the test set;

[0037] The process of randomly dividing into a training set and a test set, the specific logic is: perform a random sorting process on the sample data set, and use 80% of the sorted sample data set as the training set, and the remaining 20% as the test set.

[0038] Furthermore, the error is the mean absolute error, root mean square error, and coefficient of determination between the prediction results and the actual values in the test set, specifically as follows:

[0039] The mean absolute error is:

[0040]

[0041] The root mean square error is:

[0042]

[0043] The coefficient of determination is:

[0044]

[0045] Among them, and y i respectively represent the average value, predicted value, and actual value of the remaining service life of the pipeline corresponding to the data of the i-th test sample. n is the number of test samples in the test sample set. The actual value is the mean of the actual remaining service life values of the n groups of test samples;

[0046] When MAE ≤ ∈ MAE 、RMSE ≤ ∈ RMSE And When it indicates that the error meets the preset error threshold, the model completes training; where ∈ MAE is the preset error threshold of MAE, ∈ RMSE is the preset error threshold of RMSE, is R 2 Preset error threshold.

[0047] Furthermore, the running time, geometric parameters, and physical parameters of the pipeline to be detected are collected in real time and input into the pipeline remaining service life prediction model. The model outputs the predicted remaining service life value RUL of the pipeline to be detected, and the predicted remaining service life value is corrected using the defect evaluation coefficient to obtain the fatigue state evaluation index of the pipeline to be detected. The formula is as follows:

[0048]

[0049] In the formula, RUL′ is the fatigue state evaluation index of the pipeline to be detected, RUL is the predicted remaining service life value, is the defect evaluation coefficient, α is its preset proportionality coefficient, and α > 0, RUL ref is the reference service life of the pipeline, RUL uesd is the used service life of the pipeline;

[0050] The fatigue state evaluation index of the pipeline to be detected is compared with the preset judgment threshold. The specific logic is as follows:

[0051] When RUL′ < 0.5 * RY, it is judged that the inner wall state of the current pipeline to be detected is good, the risk is low, and the pipeline can continue to be used normally, and regular monitoring and maintenance are maintained to ensure its normal operation;

[0052] When 0.5 * RY ≤ RUL′ < RY, it is judged that the inner wall state of the current pipeline to be detected is average, there is a certain fatigue risk, and a maintenance plan needs to be considered to prevent the pipeline from deteriorating further;

[0053] When RUL′ ≥ RY, it is judged that the inner wall state of the current pipeline to be detected is poor, there is a high risk, and emergency measures should be taken immediately, including shutting down the pipeline and conducting a thorough inspection and repair to ensure safety;

[0054] In the formula, RY is the judgment threshold of the fatigue state evaluation index.

[0055] The present invention also further provides a detection device for the inner wall state of a pipeline based on multi-feature fusion. The detection device for the inner wall state of a pipeline based on multi-feature fusion is used to execute the above-mentioned detection method for the inner wall state of a pipeline based on multi-feature fusion, and includes:

[0056] A defect assessment module, which is used to divide the pipeline to be detected into several pipeline sections of equal length, collect sample images of the inner wall of each pipeline section, preprocess the sample images, use the number of pixels to characterize the defect area in the sample images, and combine the wall thickness data of the pipeline section to calculate a first evaluation coefficient. The average value of the first evaluation coefficients of all pipeline sections is calculated as the defect evaluation coefficient of the pipeline to be detected. The wall thickness data includes the original wall thickness data and real-time wall thickness data of the pipeline;

[0057] A sample data set construction module, which is used to obtain the historical operation parameters of several pipelines with the same material as the pipeline to be detected, and at the same time obtain the remaining service life of the pipeline corresponding to each set of historical operation parameters. The historical operation parameters and the corresponding remaining service life of the pipeline are mapped one by one to generate a sample data set. The historical operation parameters include operation time, geometric parameters, and environmental parameters. The geometric parameters are specifically the pipeline diameter, pipeline length, and pipeline smoothness. The environmental parameters are specifically the average temperature, average pressure, and fluid pH value;

[0058] A model construction module, which constructs a prediction model for the remaining service life of the pipeline based on a deep learning algorithm. The historical operation parameters in the sample data set are used as inputs, and the corresponding remaining service life of the pipeline is used as labels to train the prediction model for the remaining service life of the pipeline to obtain a trained prediction model for the remaining service life of the pipeline;

[0059] A fatigue state prediction module, which is used to collect the operation time, geometric parameters, and physical parameters of the pipeline to be detected in real time and input them into the prediction model for the remaining service life of the pipeline. The model outputs the predicted value of the remaining service life of the pipeline to be detected, and uses the defect evaluation coefficient to correct the predicted value of the remaining service life to obtain the fatigue state evaluation index of the pipeline to be detected;

[0060] A comprehensive evaluation module, which is used to set a judgment threshold for the fatigue state evaluation index, compare the fatigue state evaluation index of the pipeline to be detected with the threshold, and evaluate the inner wall state of the pipeline according to the comparison result to guide management decisions.

[0061] Compared with the prior art, the beneficial effects of the present invention are:

[0062] The present invention combines multiple characteristic parameters to evaluate the pipeline state, making the evaluation results more comprehensive and avoiding the limitations of single detection means. The remaining service life of the pipeline is predicted using deep learning algorithms and historical operating parameter data, making the evaluation process more scientific and capable of reflecting the fatigue state of the pipeline in real time. In addition, through the preprocessing of sample images and the accurate calculation of the defect area, combined with real-time wall thickness data, a comprehensive first evaluation coefficient is constructed, making the quantification of defects more accurate. This method can adapt to different environmental conditions to ensure the reliability of the evaluation results. These beneficial effects can effectively reduce the pipeline failure rate, extend the service life of the pipeline, provide a scientific basis for pipeline management, thereby realizing the effective utilization of resources and ensuring the safe operation of industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is a schematic diagram of the overall method flow of the present invention;

[0064] Figure 2 is a schematic diagram of the overall device module of the present invention;

[0065] Figure 3 is a comparison curve of the actual value and the predicted value of the remaining life;

[0066] Figure 4 is an accuracy rate curve of the actual value and the predicted value of the remaining life. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0067] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the following further elaborates on the present invention in conjunction with specific embodiments.

[0068] It should be noted that unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meaning understood by those of ordinary skill in the field to which the present invention belongs. The "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left" and "right" are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0069] Embodiment:

[0070] Please refer to Figure 1 , the present invention provides a technical solution:

[0071] A detection method for the inner wall state of a pipeline based on multi-feature fusion, and the specific steps include:

[0072] Step 1: Divide the pipeline to be detected into several pipeline sections of equal length, collect sample images of the inner wall of each pipeline section, after preprocessing the sample images, use the number of pixels to characterize the defect area in the sample image, and combine the wall thickness data of the pipeline section to calculate the first evaluation coefficient, and calculate the mean value of the first evaluation coefficients of all pipeline sections as the defect evaluation coefficient of the pipeline to be detected. The wall thickness data includes the original wall thickness data and the real-time wall thickness data of the pipeline;

[0073] In this embodiment, the preprocessing of the sample image includes noise removal processing and contrast enhancement processing;

[0074] Among them, the specific logic for noise removal processing is: use the denoising method of wavelet transform to denoise the sample image, decompose the image after distortion correction by wavelet transform to obtain wavelet coefficients of the image at different scales and directions; perform threshold processing on the wavelet coefficients, set the low-amplitude wavelet coefficients to zero, and retain the high-amplitude wavelet coefficients; perform inverse transform on the wavelet coefficients after threshold processing, and reconstruct the processed coefficients into an image to complete the image denoising processing;

[0075] The specific logic for contrast enhancement processing is: find the minimum gray value L max and the maximum gray value L max in the sample image to be processed, and apply the following linear transformation formula to complete the contrast enhancement processing:

[0076]

[0077] where I new (x, y) is the pixel value at the coordinate (x, y) after contrast enhancement processing, I old (x, y) refers to the original image pixel value at the coordinate (x, y), and L is the number of gray levels.

[0078] The specific logic for using the number of pixels to characterize the defect area in the sample image is: set a threshold according to the defect characteristics, and convert the gray image into a binary image, and the expression is as follows:

[0079]

[0080] where I binary (x, y) represents the value of the generated binary image at the position (x, y). If the pixel belongs to the defect area, then I binary (x, y) is 1, otherwise I binary (x, y) is 0; I gray(x, y) represents the value of the original grayscale image at the position (x, y), and T is the defect threshold set according to experiments, which is used to distinguish the defect area and the normal area in the image;

[0081] Use the connectivity algorithm to find all connected defect areas and label them. For each labeled connected area, calculate its area, which is obtained by counting the number of pixels with a value of 1 in the binary image:

[0082]

[0083] In the formula, A defect represents the number of pixels in the defect area, P(j) represents the number of pixels in the j-th defect area, j is the index of the defect area, and N is the total number of defect areas;

[0084] At the same time, obtain the spatial resolution of the image, that is, the actual area represented by each pixel. Convert the number of pixels in the defect area to the actual defect area through the following formula:

[0085] A actual = A defect * A pixel

[0086] In the formula, A actual is the actual defect area, and A pixel is the spatial resolution of the image;

[0087] Calculate the defect ratio based on the actual defect area and the total area of the pipeline section. The formula is as follows:

[0088]

[0089] In the formula, A is the defect ratio, A actual is the actual defect area, and A total is the total area of the pipeline section.

[0090] Obtain the wall thickness data of the pipeline section. Combine the defect ratio of the corresponding pipeline section. After dimensionless processing of the wall thickness data and the defect ratio, calculate the first evaluation coefficient. The formula is as follows:

[0091]

[0092] In the formula, AS is the first evaluation coefficient, which is used to characterize the overall health condition and safety of the inner wall of the pipeline. By combining the defect ratio and wall thickness data, it provides a quantitative evaluation of the state of the inner wall of the pipeline. The larger its value, the greater the impact on the inner wall of the pipeline and the higher the safety risk; A is the defect ratio, Thi0 is the original wall thickness data of the pipeline, Thi is the real-time wall thickness data, k1 and k2 are preset weights, where k1>k2>0 and k1 + k2 = 1. The reason for setting the weight values like this is that the defect ratio A is a key indicator for evaluating the state of the inner wall of the pipeline and is directly related to the safety and reliability of the pipeline. The larger the defect ratio, the higher the risk of pipeline failure. Therefore, when calculating the first evaluation coefficient, a higher weight should be given to the impact of the defect ratio to reflect its direct threat to the overall state of the pipeline. The wall thickness of the pipeline is usually a relatively stable parameter, especially when it is determined during pipeline design and manufacturing. Although the real-time wall thickness data may change due to corrosion or wear, this change is usually gradual and does not significantly affect the safety of the pipeline instantly like the defect area. Therefore, the weight of the wall thickness data in the evaluation is relatively low.

[0093] The squared term A 2 The introduction of makes the first evaluation coefficient more sensitive to changes in the defect ratio. When the actual area of the defect changes, the squared term will cause a more obvious change in the first evaluation coefficient, which is very important for timely identifying changes in the pipeline state and making maintenance decisions; when the defect ratio A increases, it means that there is a larger damaged or corroded area on the inner wall of the pipeline. Also, because the first evaluation coefficient is used to characterize the overall health condition and safety of the inner wall of the pipeline, the larger its value, the greater the impact on the inner wall of the pipeline and the higher the safety risk. Therefore, when A increases, the first evaluation coefficient AS increases accordingly; in In the term, the form of the numerator directly reflects the degree of wall thickness loss of the pipeline. Through this form, the impact of the wall thickness on the structural integrity of the pipeline can be clearly quantified. The greater the loss, the higher the numerator value, and the first evaluation coefficient also increases accordingly, reflecting an increase in the risk level of the pipeline; the denominator is set in the form of 1 + Thi0 to prevent the first evaluation index from being too sensitive, especially when the original wall thickness is large, so as to avoid misjudgment caused by an abnormally high value of the first evaluation index due to a single defect. Therefore, when the real-time wall thickness data Thi decreases, the first evaluation coefficient AS increases; that is to say, A is positively correlated with AS, and Thi is negatively correlated with AS.

[0094] The wall thickness data includes the original wall thickness data and the real-time wall thickness data of the pipeline;

[0095] Calculate the mean value of the first evaluation coefficients of all pipeline sections as the defect evaluation coefficient of the pipeline to be detected, denoted as

[0096] The advantage of Step 1 is that by equally dividing the pipeline to be detected into several sections and collecting and preprocessing sample images for each section, it is possible to achieve localized and refined detection of defects on the inner wall of the pipeline. This method not only improves the accuracy of defect detection but also enables timely identification of specific problems in different sections, facilitating the formulation of targeted maintenance plans. Compared with the prior art, this method calculates the first evaluation coefficient by combining wall thickness data, making the evaluation result more comprehensive, reflecting the comprehensive performance of the pipeline under different states, reducing the possibility of missed detection and misjudgment, and enhancing the safety and effectiveness of pipeline management.

[0097] In this solution, adopting Step 1 can provide important basic data and evaluation basis for the overall solution. Through accurate defect evaluation coefficients, the subsequent remaining service life prediction model can be trained and optimized based on more reliable data, thereby improving the accuracy of overall detection and prediction. The effective implementation of this step not only guarantees the safe operation of the pipeline but also lays a solid foundation for subsequent data analysis and decision-making support, further promoting the intelligent and scientific process of pipeline management.

[0098] Step 2: Obtain the historical operation parameters of several pipelines made of the same material as the pipeline to be detected. At the same time, obtain the remaining service life of the pipeline corresponding to each set of historical operation parameters, map the historical operation parameters and the corresponding remaining service life of the pipeline one by one to generate a sample data set. The historical operation parameters include operation time, geometric parameters, and environmental parameters. The geometric parameters are specifically the pipeline diameter, pipeline length, and pipeline smoothness. The environmental parameters are specifically the average temperature, average pressure, and fluid pH value.

[0099] Step 3: Build a pipeline remaining service life prediction model based on a deep learning algorithm. Use the historical operation parameters in the sample data set as inputs and the corresponding remaining service life of the pipeline as labels to train the pipeline remaining service life prediction model to obtain a trained pipeline remaining service life prediction model.

[0100] In this embodiment, the specific logic for building the pipeline remaining service life prediction model is as follows:

[0101] Obtain the historical operation parameters of several groups of pipelines with the same material as the pipeline to be detected. At the same time, obtain the remaining service life of the pipeline corresponding to each group of historical operation parameters. Among them, through the historical operation records of the pipeline, trace the actual failure time of the pipeline, and subtract the service life of the pipeline when it runs to failure from the total design life of the pipeline to obtain the remaining service life of the pipeline; map the historical operation parameters to the corresponding remaining service life of the pipeline one by one to generate a sample data set, and randomly divide the sample data set into a training set and a test set. The historical operation parameters include operation time, geometric parameters, and environmental parameters. The geometric parameters are specifically the pipeline diameter, pipeline length, and pipeline smoothness. The environmental parameters are specifically the average temperature, average pressure, and fluid pH value; construct a prediction model for the remaining service life of the pipeline based on a deep learning algorithm. Use the historical operation parameters in the training set as input and the corresponding remaining service life of the pipeline as labels to train the model to obtain a trained prediction model for the remaining service life of the pipeline. Substitute the historical operation parameters in the test set into the trained model to obtain the corresponding prediction results; calculate the error between the prediction results and the actual values in the test set; determine whether the error meets a preset error threshold; if it meets, output the trained model, that is, the prediction model for the remaining service life of the pipeline; if it does not meet, return to continue training; the error is the mean absolute error, root mean square error, and coefficient of determination between the prediction results and the actual values in the test set;

[0102] The pipeline smoothness is represented by the Darcy friction coefficient;

[0103] To determine the average temperature and average pressure, the specific logic is as follows: Collect the temperature inside the pipeline and the pressure on the inner wall multiple times, and take the average value as the average temperature and average pressure of the inner wall of the pipeline.

[0104] The process of randomly dividing into a training set and a test set, the specific logic is as follows: Randomly sort the sample data set, and take 80% of the sorted sample data set as the training set and the remaining 20% as the test set.

[0105] The error is the mean absolute error, root mean square error, and coefficient of determination between the prediction results and the actual values in the test set, specifically as follows:

[0106] The mean absolute error is:

[0107]

[0108] The root mean square error is:

[0109]

[0110] The coefficient of determination is:

[0111]

[0112] Among them, and y i respectively represent the average value, predicted value, and actual value of the remaining service life of the pipeline corresponding to the data of the i-th group of test samples. n is the number of groups of test samples in the test sample set, and the actual value is the mean of the actual remaining service lives of the pipelines in n groups of test samples;

[0113] When MAE ≤ ∈ MAE , RMSE ≤ ∈ RMSE and it indicates that the error meets the preset error threshold, and the model is trained; among them, ∈ MAE is the preset error threshold for MAE, ∈ RMSE is the preset error threshold for RMSE, is the preset error threshold for R 2 preset error threshold.

[0114] Table 1: Statistical Table of Historical Sample Data

[0115]

[0116]

[0117]

[0118] Data analysis: Pipeline data with several different characteristic parameters are collected, and the data range and numerical settings conform to the actual use environment of the pipeline and meet the working conditions of general industrial pipelines.

[0119] At the same time, analyze the correlation between the collected characteristic parameters and the remaining service life.

[0120] Obtain the historical operation parameters of several pipelines with the same material as the pipeline to be detected, and at the same time obtain the remaining service life of the pipeline corresponding to each group of historical operation parameters. Map the historical operation parameters to the corresponding remaining service life of the pipeline one by one to generate a sample data set. Among them, the first 20 groups of data in the sample data set are used as training group data, and the last 20 groups of data are used as verification group to detect the detection data of the life prediction model.

[0121] The prediction accuracy of the life prediction model is represented by the difference between the predicted value and the actual value of the validation group life. In this study, the characteristic parameter data of several pipelines were collected, covering aspects such as operation time, pipeline diameter, pipeline length, pipeline smoothness, average temperature, average pressure, fluid pH value, and remaining service life. The data ranges and numerical settings conform to the actual use environment of the pipelines and the working conditions of general industrial pipelines. Through correlation analysis of these characteristic parameters and the remaining service life, methods such as correlation coefficient analysis and regression analysis were used to explore the influence degree of each parameter on the pipeline service life. In addition, in order to establish an effective life prediction model, pipeline data with the same material as the pipeline to be detected were extracted from the obtained historical operation parameters, and the corresponding remaining service life was recorded, thereby generating a sample data set. The first 20 groups of data in this data set were divided into training group data, and the latter 20 groups of data were used as validation group data to detect the prediction performance of the model. After the model training was completed, the prediction accuracy was calculated by comparing the predicted value and the actual value of the validation group life.

[0122] Table 2: Statistical Table of Prediction Model Output Accuracy

[0123]

[0124]

[0125] Please refer to Figures 3 - 4, in this data analysis, the collected pipeline operation data was systematically evaluated to understand the relationship between various characteristic parameters and the remaining service life. The sample data table contains the characteristic parameters and actual remaining service life of multiple pipelines, including operation time, pipeline diameter, pipeline length, pipeline smoothness, average temperature, average pressure, fluid pH value, etc. The settings of these parameters reflect the usage of pipelines in the actual industrial environment. Through analysis, it was observed that there is a certain correlation between all characteristic parameters and the remaining service life. For example, as can be seen from the data in Table 1, parameters such as the operation time and pipeline diameter of the pipeline have a significant impact on the remaining service life. Higher average temperature and pressure are usually associated with a shorter remaining service life. In the evaluation of predicting the remaining service life, the calculated prediction accuracy shows high reliability. The vast majority of prediction accuracies are above 88%, indicating that our prediction model can effectively reflect the actual situation of the pipeline in most cases. For example, the prediction accuracy of sample number 22 reaches 93.52%, showing the rigor and effectiveness of the model in dealing with similar data. This high accuracy can enhance the monitoring and evaluation of the pipeline status, contribute to formulating scientific maintenance and management strategies, and ensure the safe and efficient operation of the pipeline. By further analyzing the differences between the predicted values and the actual values, we can gain a deeper understanding of the prediction ability of the model, providing data support and decision-making basis for future pipeline management.

[0126] The advantage of Step 2 is that by obtaining the historical operation parameters of pipelines with the same material as the pipeline to be detected and mapping these parameters to the corresponding remaining service life one by one, a scientific and systematic sample data set can be established. This data-driven method enables the evaluation of pipelines to not only rely on a single parameter but also comprehensively consider multiple geometric and environmental factors, improving the accuracy and reliability of the evaluation. Compared with the existing technology, this step avoids using outdated or incomplete empirical data, ensuring a more solid foundation for model training and effectively reducing prediction errors.

[0127] The advantage of Step 3 is that by constructing a prediction model for the remaining service life of pipelines based on deep learning algorithms, it can handle complex non-linear relationships and automatically extract features, enhancing the accuracy and generalization ability of the model. Compared with traditional linear regression or empirical models, deep learning can better adapt to the changes in multi-dimensional data, improving the prediction performance of the remaining service life of pipelines and thus enhancing the scientific nature of the decision-making process.

[0128] In this solution, adopting Steps 2 and 3 can provide strong data support and model guarantee for the overall solution. By establishing a sound sample data set and an efficient prediction model, subsequent pipeline status evaluation and management decisions will be more scientific and accurate, thereby promoting the intelligent management and maintenance of pipelines, reducing the failure rate, and ensuring the safe operation of industry. Therefore, the overall solution has higher practicality and application value.

[0129] Step 4: Real-time collect the running time, geometric parameters, and physical parameters of the pipeline to be detected, and input them into the pipeline remaining service life prediction model. The model outputs the predicted value of the remaining service life of the pipeline to be detected, and uses the defect evaluation coefficient to correct the predicted value of the remaining service life to obtain the fatigue state evaluation index of the pipeline to be detected;

[0130] In this embodiment, the running time, geometric parameters, and physical parameters of the pipeline to be detected are collected in real time and input into the pipeline remaining service life prediction model. The model outputs the predicted value of the remaining service life RUL of the pipeline to be detected, and uses the defect evaluation coefficient to correct the predicted value of the remaining service life to obtain the fatigue state evaluation index of the pipeline to be detected. The formula is as follows:

[0131]

[0132] In the formula, RUL′ is the fatigue state evaluation index of the pipeline to be detected, which is used to characterize the fatigue degree and health state of the inner wall of the pipeline. The higher its value, the higher the fatigue degree of the inner wall of the pipeline to be detected and the worse the state; RUL is the predicted value of the remaining service life. is the defect evaluation coefficient, α is its preset proportional coefficient, and 0 < α < 1. The specific value is determined according to pipeline materials, design, usage conditions, and historical fatigue performance data; RUL ref is the reference service life of the pipeline, RUL uesd is the used service life of the pipeline; among them, RUL ref and RUL uesd are obtained according to material characteristics and pipeline operation data.

[0133] In this formula, for part, is an exponential function, indicating the non-linear influence of the defect evaluation coefficient on the fatigue state evaluation index. When increases, term will increase significantly, meaning that the health state of the inner wall of the pipeline deteriorates with the increase of defects; adding a constant term +1 to this term ensures that even when the defect evaluation coefficient is 0, RUL′ still has a minimum value, maintaining the stability of the formula; for The part reflects the usage of the pipeline and is an important indicator for evaluating the health status of the inner wall of the pipeline; the logarithmic function is used to process a wide range of numerical values and can effectively balance and compress the variable RUL values. The square term Even in very large cases, the logarithmic value can avoid being too large, thus maintaining a reasonable range of the fatigue state evaluation index; adding a constant term +1 to this item ensures that the logarithmic term is not zero, ensuring the validity of the formula and avoiding division-by-zero errors.

[0134] The predicted remaining useful life value RUL output by the model is a quantitative estimate of the remaining time or cycle that the pipeline to be detected is expected to be safely used under the current operating conditions. The higher its value, the longer the expected service life of the pipeline to be detected. When RUL increases, it means that the inner wall of the pipeline has experienced less damage or defects under the current operating conditions, or the identified defects have not caused serious impacts on the overall structure of the pipeline. This means that the deterioration process of the pipeline under physical and chemical actions is slower, and the inner wall remains in a better state. Therefore, RUL' will decrease; RUL uesd The +RUL part represents the actual total life of the pipeline, represents the ratio of the reference life of the pipeline to the actual total life of the pipeline, represents the ratio of the actual total life of the pipeline to the remaining useful life of the pipeline. The smaller these two ratios are, the healthier the state of the inner wall of the pipeline is.

[0135] Due to the defect evaluation coefficient being the mean result of the first evaluation coefficient, it is also used to characterize the overall health status and safety of the inner wall of the pipeline. By combining the defect ratio and wall thickness data, it provides a quantitative evaluation of the state of the inner wall of the pipeline. The larger its value, the greater the impact on the inner wall of the pipeline and the higher the safety risk; therefore, when increases, the fatigue state evaluation index RUL' increases; that is, it indicates that is positively correlated with RUL', and RUL is negatively correlated with RUL′;

[0136] The advantage of step 4 is to collect the geometric parameters and physical parameters of the pipeline to be detected in real time and input them into the pipeline remaining useful life prediction model, making the prediction process more dynamic and flexible. This real-time nature enables the model to promptly reflect the current operating state of the pipeline, rather than relying solely on historical data, thereby improving the accuracy and timeliness of the model prediction. Compared with the existing technology, this step can more effectively capture the changes of the pipeline under different working environments by combining real-time data, reducing the risk of misjudgment caused by delay or static data, and ensuring the safety and reliability of the pipeline.

[0137] In this solution, adopting Step 4 can provide a key real-time feedback mechanism for the overall solution, improving the intelligent level of pipeline monitoring and management. By dynamically obtaining and analyzing current geometric and physical parameters, the model can calculate the predicted remaining service life more accurately and timely correct the fatigue state assessment index. This real-time adjustment can not only help detect potential problems in a timely manner, but also provide a more reliable basis for management decisions, ensuring the safe operation of the pipeline during use, thereby effectively reducing the probability of accidents and promoting the scientific and modern process of pipeline management.

[0138] Step 5: Set the judgment threshold of the fatigue state assessment index, compare the fatigue state assessment index of the pipeline to be detected with the threshold, and based on the comparison result, evaluate the inner wall state of the pipeline and guide management decisions;

[0139] In this embodiment, comparing the fatigue state assessment index of the pipeline to be detected with the preset judgment threshold, the specific logic is as follows:

[0140] When RUL′ < 0.5*RY, it is judged that the inner wall state of the current pipeline to be detected is good, the risk is low, the pipeline can continue to be used normally, and regular monitoring and maintenance are maintained to ensure its normal operation;

[0141] When 0.5*RY ≤ RUL′ < RY, it is judged that the inner wall state of the current pipeline to be detected is average, there is a certain fatigue risk, and a maintenance plan needs to be considered to prevent the pipeline from deteriorating further;

[0142] When RUL′ ≥ RY, it is judged that the inner wall state of the current pipeline to be detected is poor, there is a high risk, and emergency measures should be taken immediately, including shutting down the pipeline and conducting a thorough inspection and repair to ensure safety;

[0143] In the formula, RY is the judgment threshold of the fatigue state assessment index.

[0144] Among them, the specific logic for setting RY is as follows: Obtain examples of pipeline failures, fatigue, and corrosion from historical operation and maintenance records. Analyze these data to determine the actual performance of the pipeline under different fatigue state assessment indexes, and combine the experience of industry experts to determine the health threshold of the pipeline under specific materials and specific operating conditions.

[0145] The advantage of Step 5 is that by setting the judgment threshold of the fatigue state assessment index and comparing it with the actual fatigue state assessment index, a systematic evaluation of the inner wall state of the pipeline can be achieved. This method provides a clear basis for management decisions, making the evaluation results more intuitive and operable. Compared with the existing technology, this step not only improves the accuracy of pipeline monitoring, but also enhances the ability to respond to potential risks, making maintenance and management measures more timely and effective, thereby reducing safety hazards caused by pipeline failures.

[0146] In this solution, adopting Step 5 can provide crucial decision-making support for the overall solution, ensuring the safety and stability of the pipeline at different usage stages. Through clear threshold comparison, managers can quickly identify the specific state of the pipeline and take appropriate maintenance measures according to the results. This systematic evaluation not only improves the scientific nature and efficiency of pipeline management but also promotes the intelligent process of pipeline monitoring and maintenance, driving the improvement of the overall pipeline management level and ensuring the safety and economy of operation.

[0147] Please refer to Figure 2 , a detection device for the inner wall state of a pipeline based on multi-feature fusion, comprising:

[0148] A defect evaluation module, which is used to divide the pipeline to be detected into several pipeline sections of equal length, collect sample images of the inner wall of each pipeline section, preprocess the sample images, use the number of pixels to characterize the defect area in the sample images, and combine the wall thickness data of the pipeline section to calculate the first evaluation coefficient. Calculate the mean value of the first evaluation coefficients of all pipeline sections as the defect evaluation coefficient of the pipeline to be detected. The wall thickness data includes the original wall thickness data and the real-time wall thickness data of the pipeline;

[0149] A sample data set construction module, which is used to obtain the historical operation parameters of several pipelines with the same material as the pipeline to be detected, and at the same time obtain the remaining service life of the pipeline corresponding to each set of historical operation parameters, map the historical operation parameters to the corresponding remaining service life of the pipeline one by one to generate a sample data set. The historical operation parameters include operation time, geometric parameters, and environmental parameters. The geometric parameters are specifically the pipeline diameter, pipeline length, and pipeline smoothness, and the environmental parameters are specifically the average temperature, average pressure, and fluid pH value;

[0150] A model construction module, which constructs a prediction model for the remaining service life of the pipeline based on a deep learning algorithm, uses the historical operation parameters in the sample data set as inputs, and uses the corresponding remaining service life of the pipeline as labels to train the prediction model for the remaining service life of the pipeline to obtain a trained prediction model for the remaining service life of the pipeline;

[0151] A fatigue state prediction module, which is used to collect the operation time, geometric parameters, and physical parameters of the pipeline to be detected in real time and input them into the prediction model for the remaining service life of the pipeline. The model outputs the predicted value of the remaining service life of the pipeline to be detected, and uses the defect evaluation coefficient to correct the predicted value of the remaining service life to obtain the fatigue state evaluation index of the pipeline to be detected;

[0152] The comprehensive evaluation module is used to set the judgment threshold of the fatigue state evaluation index, compare the fatigue state evaluation index of the pipeline to be detected with the threshold, evaluate the inner wall state of the pipeline according to the comparison result, and guide the management decision-making.

[0153] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0154] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0155] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0156] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application.

Claims

1. A method for detecting the inner wall state of a pipeline based on multi-feature fusion, characterized in that: The specific steps include: Step 1: Divide the pipeline to be inspected into several pipeline sections of equal length, collect sample images of the inner wall of each pipeline section, pre-process the sample images, use the number of pixels to represent the defect area in the sample images, and calculate the first evaluation coefficient in combination with the wall thickness data of the pipeline section, and calculate the average of the first evaluation coefficients of all pipeline sections as the defect evaluation coefficient of the pipeline to be inspected, wherein the wall thickness data includes the original wall thickness data and the real-time wall thickness data of the pipeline; Step 2: Obtain several groups of historical operating parameters of pipelines of the same material as the pipeline to be tested, and simultaneously obtain the remaining service life of the pipeline corresponding to each group of historical operating parameters, and map the historical operating parameters to the corresponding remaining service life of the pipeline one by one to generate a sample data set, wherein the historical operating parameters include operating time, geometric parameters and environmental parameters, wherein the geometric parameters are specifically the pipeline diameter, pipeline length and pipeline smoothness, and the environmental parameters are specifically the average temperature, average pressure and fluid pH value; Step 3: Build a pipeline remaining service life prediction model based on the deep learning algorithm, take the historical operating parameters in the sample data set as input, take the corresponding pipeline remaining service life as the label, train the pipeline remaining service life prediction model, and obtain a trained pipeline remaining service life prediction model; Step 4: The operating time, geometric parameters and physical parameters of the pipeline to be inspected are collected in real time and input into the pipeline remaining service life prediction model. The model outputs the remaining service life prediction value of the pipeline to be inspected. The remaining service life prediction value is corrected by using the defect assessment coefficient to obtain the fatigue state assessment index of the pipeline to be inspected. Step 5: Set the judgment threshold of the fatigue status assessment index, compare the fatigue status assessment index of the pipeline to be inspected with the threshold, and evaluate the inner wall status of the pipeline based on the comparison result to guide management decisions.

2. The method for detecting the inner wall state of a pipeline based on multi-feature fusion according to claim 1, characterized in that: The preprocessing of the sample image includes noise removal and contrast enhancement; The specific logic of noise removal is as follows: the sample image is denoised using the wavelet transform denoising method, the distortion-corrected image is decomposed by wavelet transform to obtain the wavelet coefficients of the image at different scales and directions; the wavelet coefficients are thresholded to set the low-amplitude wavelet coefficients to zero and retain the high-amplitude wavelet coefficients; the wavelet coefficients after thresholding are inversely transformed to reconstruct the processed coefficients into an image to complete the image denoising; The specific logic of contrast enhancement processing is: find the minimum gray value L in the sample image to be processed max and the maximum gray value L max , the following linear transformation formula is applied to complete the contrast enhancement process: In the formula, I new (x, y) is the pixel value at coordinate (x, y) after contrast enhancement processing, I old (x,y) refers to the original image pixel value at coordinate (x,y), and L is the number of gray levels.

3. The method for detecting the inner wall state of a pipeline based on multi-feature fusion according to claim 1, characterized in that: The specific logic of using the number of pixels to characterize the defect area in the sample image is: setting a threshold according to the defect characteristics and converting the grayscale image into a binary image. The expression is as follows: In the formula, I binary (x, y) represents the value of the generated binary image at the position (x, y). If the pixel belongs to the defect area, I binary (x,y) is 1, otherwise I binary (x,y) is 0; I gray (x, y) represents the value of the original grayscale image at the position (x, y), T is the defect threshold set according to the experiment, which is used to distinguish the defect area and the normal area in the image; Use the connectivity algorithm to find all connected defect areas and mark them, and calculate the area of ​​each marked connected area. The area is obtained by counting the number of pixels with a value of 1 in the binary image: In the formula, A defect represents the number of pixels in the defective area, P(j) represents the number of pixels in the jth defective area, j is the index of the defective area, and N is the total number of defective areas; At the same time, the spatial resolution of the image is obtained, that is, the actual area represented by each pixel, and the number of pixels in the defect area is converted into the actual defect area by the following formula: A actual =A defect *A pixel In the formula, A actual is the actual defect area, A pixel is the spatial resolution of the image; The defect ratio is calculated based on the actual defect area and the total area of ​​the pipeline section according to the following formula: Where A is the defect ratio, A actual is the actual defect area, A total is the total area of ​​the pipeline section.

4. The method for detecting the inner wall state of a pipeline based on multi-feature fusion according to claim 1, characterized in that: Obtain the wall thickness data of the pipeline section, combine it with the defect ratio of the corresponding pipeline section, and calculate the first evaluation coefficient after dimensionless processing of the wall thickness data and the defect ratio. The formula is as follows: Where AS is the first evaluation coefficient, A is the defect ratio, Thi0 is the original wall thickness data of the pipeline, Thi is the real-time wall thickness data, k1 and k2 are preset weights, k1>k2>0, and k1+k2=1; The wall thickness data includes original pipe wall thickness data and real-time pipe wall thickness data; The first evaluation coefficients of all pipeline sections are averaged and used as the defect evaluation coefficient of the pipeline to be inspected, denoted as 5. The method for detecting the inner wall state of a pipeline based on multi-feature fusion according to claim 1, characterized in that: The specific logic of constructing the pipeline remaining service life prediction model is as follows: Obtain several groups of historical operating parameters of pipelines with the same material as the pipeline to be tested, and at the same time obtain the remaining service life of the pipeline corresponding to each group of historical operating parameters, map the historical operating parameters with the corresponding remaining service life of the pipeline one by one, generate a sample data set, and randomly divide the sample data set into a training set and a test set, wherein the historical operating parameters include operating time, geometric parameters and environmental parameters, the geometric parameters are specifically the pipeline diameter, pipeline length and pipeline smoothness, and the environmental parameters are specifically the average temperature, average pressure and fluid pH value; construct a pipeline remaining service life prediction model based on a deep learning algorithm, take the historical operating parameters in the training set as input, take the corresponding remaining service life of the pipeline as a label, train the model to obtain a trained pipeline remaining service life prediction model, substitute the historical operating parameters in the test set into the trained model, and obtain the corresponding prediction result; calculate the error between the prediction result and the actual value in the test set; determine whether the error meets the preset error threshold; if so, output the trained model, that is, the pipeline remaining service life prediction model; if not, return to continue training; the error is the mean absolute error, root mean square error and determination coefficient between the prediction result and the actual value in the test set; The process of randomly dividing into training sets and test sets has the following specific logic: randomly sort the sample data set, use 80% of the sorted sample data set as the training set, and the remaining 20% ​​as the test set.

6. The method for detecting the inner wall state of a pipeline based on multi-feature fusion according to claim 5, characterized in that: The error is the mean absolute error, root mean square error and determination coefficient between the predicted result and the actual value in the test set, as follows: The mean absolute error is: The root mean square error is: The coefficient of determination is: in, and i They respectively represent the average value, predicted value and actual value of the remaining service life of the pipeline corresponding to the data of the i-th group of test samples, n is the number of test sample groups in the test sample set, and the actual value is the average value of the actual values ​​of the remaining service life of the pipeline of n groups of test samples; When MAE≤∈ MAE , RMSE≤∈ RMSE and , it means that the error meets the preset error threshold and the model has completed training; where ∈ MAE Preset error threshold for MAE, ∈ RMSE Preset error threshold for RMSE, For R 2 Preset error threshold.

7. The method for detecting the inner wall state of a pipeline based on multi-feature fusion according to claim 1, characterized in that: The running time, geometric parameters and physical parameters of the pipeline to be inspected are collected in real time and input into the pipeline remaining service life prediction model. The model outputs the remaining service life prediction value RUL of the pipeline to be inspected. The remaining service life prediction value is corrected by using the defect assessment coefficient to obtain the fatigue state assessment index of the pipeline to be inspected. The formula is as follows: In the formula, RUL′ is the fatigue status evaluation index of the pipeline to be inspected, RUL is the remaining service life prediction value, is the defect assessment coefficient, α is its preset proportional coefficient, and α>0, RUL ref is the reference service life of the pipeline, RUL uesd is the service life of the pipeline; The fatigue status evaluation index of the pipeline to be inspected is compared with the preset judgment threshold, and the specific logic is as follows: When RUL′<0.5*RY, it is judged that the inner wall of the pipeline to be inspected is in good condition and the risk is low. The pipeline should continue to be used normally and regular monitoring and maintenance should be maintained to ensure its normal operation. When 0.5*RY ≤ RUL′ < RY, it is judged that the inner wall state of the pipeline to be detected is general at present, there is a certain fatigue risk, and a maintenance plan needs to be considered to prevent the pipeline from deteriorating further; When RUL′ ≥ RY, it is judged that the inner wall state of the pipeline to be detected is poor and there is a high risk. Immediate emergency measures should be taken, including shutting down the pipeline and conducting a thorough inspection and repair to ensure safety; In the formula, RY is the judgment threshold of the fatigue state evaluation index.

8. A detection device for the inner wall state of a pipeline based on multi-feature fusion, characterized in that: The described detection device for the inner wall state of a pipeline based on multi-feature fusion is used to execute the detection method for the inner wall state of a pipeline based on multi-feature fusion according to any one of claims 1-7, including: A defect evaluation module, which is used to divide the pipeline to be detected into several pipeline sections of equal length, collect sample images of the inner wall of each pipeline section, preprocess the sample images, use the number of pixels to represent the defect area in the sample images, and combine the wall thickness data of the pipeline section to calculate the first evaluation coefficient. Calculate the mean value of the first evaluation coefficients of all pipeline sections as the defect evaluation coefficient of the pipeline to be detected. The wall thickness data includes the original wall thickness data and the real-time wall thickness data of the pipeline; A sample data set construction module, which is used to obtain the historical operation parameters of several pipelines with the same material as the pipeline to be detected, and at the same time obtain the remaining service life of the pipeline corresponding to each set of historical operation parameters, map the historical operation parameters and the corresponding remaining service life of the pipeline one by one to generate a sample data set. The historical operation parameters include geometric parameters and environmental parameters. The geometric parameters are specifically the pipeline diameter, pipeline length and pipeline smoothness, and the environmental parameters are specifically the average temperature, average pressure and fluid pH value; A model construction module, which constructs a pipeline remaining service life prediction model based on a deep learning algorithm, uses the historical operation parameters in the sample data set as inputs, and uses the corresponding remaining service life of the pipeline as labels to train the pipeline remaining service life prediction model to obtain a trained pipeline remaining service life prediction model; A fatigue state prediction module, which is used to collect the geometric parameters and physical parameters of the pipeline to be detected in real time and input them into the pipeline remaining service life prediction model. The model outputs the predicted value of the remaining service life of the pipeline to be detected, and uses the defect evaluation coefficient to correct the predicted value of the remaining service life to obtain the fatigue state evaluation index of the pipeline to be detected; A comprehensive evaluation module, which is used to set the judgment threshold of the fatigue state evaluation index, compare the fatigue state evaluation index of the pipeline to be detected with the threshold, and evaluate the inner wall state of the pipeline according to the comparison result and guide the management decision-making.

Images