Intelligent evaluation method and device for pipeline integrity based on detection signal

Abstract

The application discloses a kind of pipeline integrity intelligent evaluation method and device based on detection signal, method includes: obtaining the magnetic induction detection signal data of pipeline;Magnetic induction detection signal data is input into integrity intelligent evaluation model, and the integrity intelligent evaluation result of pipeline is obtained;Wherein, integrity intelligent evaluation result includes abnormal area position and repair level;Integrity intelligent evaluation model is that sample magnetic induction detection signal data is regarded as the input of model, and the abnormal area position and repair level corresponding to the sample magnetic induction detection signal data are regarded as the output of model, and the neural network model is trained and tested to obtain.This application is based on magnetic induction detection signal data, proposes a kind of pipeline integrity intelligent evaluation method with strong adaptability and not influenced by artificial subjective factors, improves the accuracy of detection result analysis, avoids blind repair, can greatly improve economy and reduce labor force, and simultaneously improves the safety of pipeline.

Description

Technical Field

[0001] This invention relates to the field of pipeline safety evaluation technology, specifically to a method and apparatus for intelligent evaluation of pipeline integrity based on detection signals. Background Technology

[0002] Pipelines within a plant are the lifeline of production in refineries and stations, serving as the link between various facilities. Inspecting buried pipelines within plants is challenging, they are severely corroded, and leaks are frequent, causing significant environmental, safety, and economic losses. Currently, the primary method for excavating and inspecting buried pipelines within plants is magnetic stress testing.

[0003] Existing magnetic induction detection signal interpretation technology mainly relies on human experience to judge the signal graph and classify the abnormality of magnetic stress detection signal according to relevant grading standards. However, the disadvantages of this method are that it introduces too many subjective human factors, the accuracy of the evaluation cannot be guaranteed, and the labor cost is high, resulting in poor economic efficiency. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention proposes a method and apparatus for intelligent evaluation of pipeline integrity based on detection signals.

[0005] Specifically, the embodiments of the present invention provide the following technical solutions:

[0006] In a first aspect, embodiments of the present invention provide a method for intelligent evaluation of pipeline integrity based on detection signals, including:

[0007] Acquire magnetic induction detection signal data of the pipeline;

[0008] The magnetic induction detection signal data is input into the integrity intelligent evaluation model to obtain the integrity intelligent evaluation result of the pipeline; wherein, the integrity intelligent evaluation result includes the location of abnormal areas and the maintenance level;

[0009] The integrity intelligent evaluation model is obtained by training and testing a neural network model, using sample magnetic induction detection signal data as input and the location and maintenance level of the abnormal area corresponding to the sample magnetic induction detection signal data as output.

[0010] Furthermore, the intelligent pipeline integrity evaluation method based on detection signals also includes: the training process of the intelligent integrity evaluation model;

[0011] The training process of the integrity intelligent evaluation model includes:

[0012] Collect a predetermined number of sample magnetic induction detection signal data; wherein, the predetermined number of sample magnetic induction detection signal data needs to cover various abnormal area locations and various maintenance levels;

[0013] Based on the experience or excavation, experts in signal analysis determine the location and maintenance level of the pipeline abnormality area corresponding to the preset number of sample magnetic induction detection signal data.

[0014] The sample magnetic induction detection signal data is used as the input to the model, and the abnormal area location and maintenance level corresponding to the sample magnetic induction detection signal data are used as the output of the model. The neural network model is trained and tested based on the intelligent learning algorithm to obtain the integrity intelligent evaluation model.

[0015] Furthermore, after collecting a preset number of sample magnetic induction detection signal data, the method further includes the following steps:

[0016] The sample magnetic induction detection signal data undergoes length uniformization processing, specifically including:

[0017] The magnetic induction detection signal for each section of the pipeline is typically a two-dimensional array, in the form of:

[0018]

[0019] Where m is the number of samplings, and the magnetic induction detection signal obtained at each sampling point is:

[0020]

[0021] Where n is the dimension of the magnetic induction detection signal, h ij This represents the numerical value of the j-th signal feature in the magnetic induction detection signal;

[0022] The sampling frequency of the magnetic induction detection signal data is uniform, and the length of H represents the length of the detected pipe segment. The magnetic induction detection signal data is standardized so that the length of each detection signal vector is the same, so as to facilitate network training.

[0023] First, the average length of the magnetic induction detection signal data for all inspected pipe sections is calculated. Round to the nearest integer, then proceed with cases 1 through 5 below:

[0024] Case 1: Regarding array length The signal data is scaled down to the desired array length using interpolation. Specifically, first obtain the positions of the sampling points of the new array in the original numerical coordinates:

[0025]

[0026] In the formula, X i This is a set of positional information for the new array after the transformation.

[0027] Based on the new location information Xi The detection signal at the new position is calculated using linear interpolation. The interpolation method is as follows:

[0028]

[0029] In the formula, j = [X i ] is X i The integer part, The new magnetic induction detection signal data H is obtained from the magnetic induction detection signal data at the corresponding position in the original data. * The vector length is

[0030] Case 2: Regarding array length Signal data, for The array is concatenated to obtain a new signal:

[0031]

[0032] The signal length becomes N*l, until... Then, following the method in case 1, scale the length to...

[0033] Case 3: Regarding array length The data without abnormal signals is divided from the start and end positions of the array towards the opposite side. The length yields two new data points:

[0034]

[0035] Case 4: Regarding array length The abnormal signal data H is obtained from the array. Starting position Cut from length 1 to obtain length 2. The signal, until the remaining length Then, following the methods in cases 1-3, change the length of the remaining part to...

[0036] Case 5: Regarding array length If there is abnormal signal data H, then the abnormal region set G is expressed as:

[0037] G = {g i ,g i +D i ], i = 1, 2, ..., q}

[0038] In the formula, q represents the number of abnormal regions in H, and g i Let D be the starting position of the i-th anomaly. iLet the location length of the i-th anomaly be denoted by . First, for each anomaly region, divide it outwards from the start and end points of the anomaly signal. Length, if or If an interval completely contains other abnormal signal intervals, then the abnormal intervals are merged and pruned. Specifically: when At that time, abnormal regions are merged into a single signal and pruned; similarly, when... At that time, abnormal regions are merged into a single signal and pruned, resulting in an array length of [length missing]. like or The interval does not completely contain other abnormal signals, from the abnormal signal [g i ,g i +D i Extending from the midpoint of the graph to both sides, we obtain a length of... interval If this interval contains endpoints of other abnormal signals, then that endpoint is used as a clipping point, and the interval is extended to the other side for clipping, ensuring that the length of the portion containing the abnormal signal is l. For the clipped interval... If the right endpoint falls within another abnormal interval at this point, then the clipping interval is further narrowed to the left until there are no abnormalities, resulting in the final clipping interval [g]. i-1 +D i-1 ,g i+1 For the final trimmed signal array, repeat cases 1-2; for The signal directly The abnormal signal region is trimmed into a signal array, and then the remaining part is processed according to cases 1-5; for the remaining no-signal region after the above trimming, cases 1-4 are executed in a loop.

[0039] Furthermore, the sample magnetic induction detection signal data is used as the input to the model, and the location of the abnormal area and the maintenance level corresponding to the sample magnetic induction detection signal data are used as the output of the model. The neural network model is trained and tested based on an intelligent learning algorithm to obtain an integrity intelligent evaluation model, including:

[0040] Based on the statistical results of historically acquired magnetic induction detection signal data, the ratio of the number of magnetic induction detection signal data containing defects to the number of magnetic induction detection signal data not containing defects is determined.

[0041] Based on the aforementioned quantity ratio, the sample magnetic induction detection signals of the pipeline anomaly area location and maintenance level are randomly divided into training set and test set to train and test the integrity intelligent evaluation model.

[0042] Furthermore, before inputting the magnetic induction detection signal data into the integrity intelligent evaluation model, the method further includes:

[0043] When applying the model, for the magnetic induction detection signal data, the array length ratio of the magnetic induction detection signal data is first standardized.

[0044] If the signal length Then, scale the signal array to the standard size using the methods described in cases 1 and 2 above;

[0045] If the signal length Then, starting from the left endpoint of the signal array, establish a length of... The sliding window, each The signal is trimmed once by length until the right side of the sliding window reaches the boundary of the signal array, resulting in... A number of signal arrays are input into a network for prediction. Based on the network prediction results, the range of abnormal signal locations in the data and the maintenance level of the abnormal signal are output.

[0046] Furthermore, the intelligent pipeline integrity evaluation method based on detection signals also includes: a process of adjusting and migrating the intelligent integrity evaluation model;

[0047] The process of adjusting and migrating the integrity intelligent evaluation model includes:

[0048] After accumulating pipeline excavation data, environmental factors affecting magnetic induction, including metal, tree, and pipeline properties, are encoded using one-hot or continuous encoding to obtain a first feature vector. The feature vector corresponding to the target region selected using a Region Generation Network (RPN) on the feature map obtained from a convolutional neural network is concatenated with the first feature vector to obtain a second feature vector. This second feature vector is then processed by a fully connected neural network layer and passed through a softmax function to calculate the probability L of the sample magnetic induction detection signal data belonging to each maintenance level. i A fully connected network is trained using the stochastic gradient descent (SGD) method, and the network output L. i In the results, the maximum L i The location indicates the most likely repair level for the abnormality.

[0049] Wherein, the probability L of a signal sample belonging to each maintenance level is... i The calculation method is as follows:

[0050]

[0051] Where, p i It is the output of the fully connected network layer, and N represents the number of maintenance levels.

[0052] Furthermore, the intelligent pipeline integrity evaluation method based on detection signals also includes: optimizing the intelligent integrity evaluation model so that the intelligent integrity evaluation model outputs defect types;

[0053] A multi-label Sigmoid loss function layer is added after the fully connected layer of the network to obtain the probability of predicting the sample magnetic induction detection signal data as each type of defect.

[0054]

[0055] in, Let M be the column vector output by the last fully connected layer of the network, where M is the number of defect types.

[0056] The network is trained using stochastic gradient descent (SGD) with cross-entropy loss function, and the network output is... The defect type corresponding to all positions in the vector that are greater than the threshold;

[0057] After completing the above processing, the network outputs the location of the abnormal area, the type of defect, and the corresponding maintenance level during prediction.

[0058] Secondly, embodiments of the present invention also provide a pipeline integrity intelligent evaluation device based on detection signals, comprising:

[0059] The acquisition module is used to acquire magnetic induction detection signal data of the pipeline;

[0060] The intelligent evaluation module is used to input the magnetic induction detection signal data into the integrity intelligent evaluation model to obtain the integrity intelligent evaluation result of the pipeline; wherein, the integrity intelligent evaluation result includes the location of abnormal areas and the maintenance level;

[0061] The integrity intelligent evaluation model is obtained by training and testing a neural network model, using sample magnetic induction detection signal data as input and the location and maintenance level of the abnormal area corresponding to the sample magnetic induction detection signal data as output.

[0062] Furthermore, the intelligent pipeline integrity evaluation device based on detection signals also includes: a training module for performing a training process on the intelligent integrity evaluation model;

[0063] The training process of the integrity intelligent evaluation model includes:

[0064] Collect a predetermined number of sample magnetic induction detection signal data; wherein, the predetermined number of sample magnetic induction detection signal data needs to cover various abnormal area locations and various maintenance levels;

[0065] Based on the experience or excavation, experts in signal analysis determine the location and maintenance level of the pipeline abnormality area corresponding to the preset number of sample magnetic induction detection signal data.

[0066] The sample magnetic induction detection signal data is used as the input to the model, and the abnormal area location and maintenance level corresponding to the sample magnetic induction detection signal data are used as the output of the model. The neural network model is trained and tested based on the intelligent learning algorithm to obtain the integrity intelligent evaluation model.

[0067] Furthermore, after collecting a preset number of sample magnetic induction detection signal data, the training module is also used to perform the following steps:

[0068] The sample magnetic induction detection signal data undergoes length uniformization processing, specifically including:

[0069] The magnetic induction detection signal for each section of the pipeline is typically a two-dimensional array, in the form of:

[0070]

[0071] Where m is the number of samplings, and the magnetic induction detection signal obtained at each sampling point is:

[0072]

[0073] Where n is the dimension of the magnetic induction detection signal, h ij This represents the numerical value of the j-th signal feature in the magnetic induction detection signal;

[0074] The sampling frequency of the magnetic induction detection signal data is uniform, and the length of H represents the length of the detected pipe segment. The magnetic induction detection signal data is standardized so that the length of each detection signal vector is the same, so as to facilitate network training.

[0075] First, the average length of the magnetic induction detection signal data for all inspected pipe sections is calculated. Round to the nearest integer, then proceed with cases 1 through 5 below:

[0076] Case 1: Regarding array length The signal data is scaled down to the desired array length using interpolation. Specifically, first obtain the positions of the sampling points of the new array in the original numerical coordinates:

[0077]

[0078] In the formula, X i This is a set of positional information for the new array after the transformation.

[0079] Based on the new location information X i The detection signal at the new position is calculated using linear interpolation. The interpolation method is as follows:

[0080]

[0081] In the formula, j = [X i ] is X i The integer part, The new magnetic induction detection signal data H is obtained from the magnetic induction detection signal data at the corresponding position in the original data. * The vector length is

[0082] Case 2: Regarding array length Signal data, for The array is concatenated to obtain a new signal:

[0083]

[0084] The signal length becomes N*l, until... Then, following the method in case 1, scale the length to...

[0085] Case 3: Regarding array length The data without abnormal signals is divided from the start and end positions of the array towards the opposite side. The length yields two new data points:

[0086]

[0087] Case 4: Regarding array length The abnormal signal data H is obtained from the array. Starting position Cut from length 1 to obtain length 2. The signal, until the remaining length Then, following the methods in cases 1-3, change the length of the remaining part to...

[0088] Case 5: Regarding array length If there is abnormal signal data H, then the abnormal region set G is expressed as:

[0089] G = {g i ,g i +D i ], i = 1, 2, ..., q}

[0090] In the formula, q represents the number of abnormal regions in H, and g i Let D be the starting position of the i-th anomaly. i Let the location length of the i-th anomaly be denoted by . First, for each anomaly region, divide it outwards from the start and end points of the anomaly signal. Length, if or If an interval completely contains other abnormal signal intervals, then the abnormal intervals are merged and pruned. Specifically: when At that time, abnormal regions are merged into a single signal and pruned; similarly, when... At that time, abnormal regions are merged into a single signal and pruned, resulting in an array length of [length missing]. like or The interval does not completely contain other abnormal signals, from the abnormal signal [g i ,g i +D i Extending from the midpoint of the graph to both sides, we obtain a length of... interval If this interval contains endpoints of other abnormal signals, then that endpoint is used as a clipping point, and the interval is extended to the other side to ensure that the length of the portion containing the abnormal signal is [length missing]. For clipping intervals If the right endpoint falls within another abnormal interval at this point, then the clipping interval is further narrowed to the left until there are no abnormalities, resulting in the final clipping interval [g]. i-1 +D i-1 ,g i+1 For the final trimmed signal array, repeat cases 1-2; for The signal directly The abnormal signal region is trimmed into a signal array, and then the remaining part is processed according to cases 1-5; for the remaining no-signal region after the above trimming, cases 1-4 are executed in a loop.

[0091] As can be seen from the above technical solution, the intelligent pipeline integrity evaluation method and device based on detection signals provided in this embodiment of the invention first acquires magnetic induction detection signal data of the pipeline, and then inputs the magnetic induction detection signal data into an intelligent integrity evaluation model to obtain the intelligent integrity evaluation result of the magnetic induction detection signal data. The intelligent integrity evaluation result includes the location of abnormal areas and the maintenance level. The intelligent integrity evaluation model is obtained by training and testing a neural network model using sample magnetic induction detection signal data with pre-determined abnormal area locations and maintenance levels as input, and the abnormal area locations and maintenance levels in the sample magnetic induction detection signal data as output. Based on magnetic induction detection signal data, this embodiment of the invention proposes a highly adaptable intelligent pipeline integrity evaluation method that is unaffected by subjective human factors, improving the accuracy of detection result analysis, avoiding blind maintenance, greatly improving economy and reducing labor costs, and simultaneously improving pipeline safety. Attached Figure Description

[0092] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0093] Figure 1 This is a flowchart of a pipeline integrity intelligent evaluation method based on detection signals provided in an embodiment of the present invention;

[0094] Figure 2 This is a schematic diagram of a pipeline integrity evaluation process based on magnetic induction detection signal data and neural network provided in an embodiment of the present invention;

[0095] Figure 3 This is a schematic diagram of the structure of an intelligent pipeline integrity evaluation device based on detection signals provided in an embodiment of the present invention;

[0096] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0097] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. The following examples are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0098] Figure 1 A flowchart of a pipeline integrity intelligent evaluation method based on detection signals provided by an embodiment of the present invention is shown, as follows: Figure 1As shown in the figure, the intelligent pipeline integrity evaluation method based on detection signals provided in this embodiment of the invention specifically includes the following:

[0099] Step 101: Acquire the magnetic induction detection signal data of the pipeline;

[0100] In this step, the magnetic induction detection signal data is a two-dimensional array signal.

[0101] Step 102: Input the magnetic induction detection signal data into the integrity intelligent evaluation model to obtain the integrity intelligent evaluation result of the pipeline; wherein, the integrity intelligent evaluation result includes the location of abnormal areas and the maintenance level;

[0102] The integrity intelligent evaluation model is obtained by training and testing a neural network model, using sample magnetic induction detection signal data as input and the location and maintenance level of the abnormal area corresponding to the sample magnetic induction detection signal data as output.

[0103] In this embodiment, it should be noted that, by utilizing magnetic induction detection signal data, a highly adaptable and unaffected intelligent evaluation method for pipeline integrity is proposed. This method improves the accuracy of the detection result analysis, avoids blind maintenance, greatly improves economic efficiency and reduces labor costs, and enhances pipeline safety.

[0104] In this embodiment, it should be noted that the pipeline detection signal data is generally a data pattern of signal strength corresponding to the pipeline mileage. This data is usually displayed and analyzed in the form of signal graphs. For example, based on the intensity value of the obtained magnetic stress signal, a corresponding magnetic stress signal image can be drawn, and then an intelligent algorithm can be used to predict the severity of pipeline defects based on the signal image. However, since digital signals lose some accuracy when converted into images, the accuracy of the algorithm is limited. Furthermore, processing and calculating images requires a large amount of computing resources, is slow, and places very high demands on computer hardware. Therefore, this embodiment proposes a pipeline integrity intelligent evaluation method that directly targets data signals. This method is not only applicable to the intelligent evaluation of pipeline integrity based on signal data from non-contact pipeline detection, such as magnetic stress detection, but can also be used for the direct intelligent evaluation of other types of detection data. On the one hand, it has a wide range of applicability; on the other hand, while improving the accuracy and efficiency of pipeline integrity evaluation, it reduces labor costs, improves economic benefits and pipeline safety, and fills the technical gap in existing intelligent pipeline integrity evaluation.

[0105] As can be seen from the above technical solution, the intelligent pipeline integrity evaluation method based on detection signals provided in this embodiment of the invention first acquires the magnetic induction detection signal data of the pipeline, and then inputs the magnetic induction detection signal data into the intelligent integrity evaluation model to obtain the intelligent integrity evaluation result of the magnetic induction detection signal data. The intelligent integrity evaluation result includes the location of abnormal areas and the maintenance level. The intelligent integrity evaluation model is obtained by training and testing a neural network model using sample magnetic induction detection signal data with pre-determined abnormal area locations and maintenance levels as input, and the abnormal area locations and maintenance levels in the sample magnetic induction detection signal data as output. This embodiment of the invention proposes an intelligent pipeline integrity evaluation method based on magnetic induction detection signal data that is highly adaptable and unaffected by subjective human factors. This improves the accuracy of detection result analysis, avoids blind maintenance, greatly improves economy and reduces labor costs, and simultaneously enhances pipeline safety.

[0106] Based on the above embodiments, in this embodiment, the intelligent evaluation method for pipeline integrity based on detection signals further includes: the training process of the intelligent evaluation model for integrity.

[0107] The training process of the integrity intelligent evaluation model includes:

[0108] Collect a predetermined number of sample magnetic induction detection signal data; wherein, the predetermined number of sample magnetic induction detection signal data needs to cover various abnormal area locations and various maintenance levels;

[0109] Based on the experience or excavation, experts in signal analysis determine the location and maintenance level of the pipeline abnormality area corresponding to the preset number of sample magnetic induction detection signal data.

[0110] The sample magnetic induction detection signal data with pre-determined abnormal area locations and maintenance levels are used as input, and the abnormal area locations and maintenance levels in the sample magnetic induction detection signal data are used as output. The neural network model is trained and tested based on the intelligent learning algorithm to obtain the integrity intelligent evaluation model.

[0111] In this embodiment, the preset number of sample magnetic induction detection signal data covers various abnormal area locations and various maintenance levels, thereby enabling the integrity intelligent evaluation model obtained after model training using the preset number of sample magnetic induction detection signal data to accurately identify the abnormal area locations and maintenance levels of the pipeline's magnetic induction detection signal data.

[0112] Based on the above embodiments, in this embodiment, after collecting a preset number of sample magnetic induction detection signal data, the method further includes the following steps:

[0113] The sample magnetic induction detection signal data undergoes length uniformization processing, specifically including:

[0114] The magnetic induction detection signal for each section of the pipeline is typically a two-dimensional array, in the form of:

[0115]

[0116] Where m is the number of samplings, and the magnetic induction detection signal obtained at each sampling point is:

[0117]

[0118] Where n is the dimension of the magnetic induction detection signal, h ij This represents the numerical value of the j-th signal feature in the magnetic induction detection signal;

[0119] The sampling frequency of the magnetic induction detection signal data is uniform, and the length of H represents the length of the detected pipe segment. The magnetic induction detection signal data is standardized so that the length of each detection signal vector is the same, so as to facilitate network training.

[0120] First, the average length of the magnetic induction detection signal data for all inspected pipe sections is calculated. Round to the nearest integer, then proceed with cases 1 through 5 below:

[0121] Case 1: Regarding array length The signal data is scaled down to the desired array length using interpolation. Specifically, first obtain the positions of the sampling points of the new array in the original numerical coordinates:

[0122]

[0123] In the formula, X i This is a set of positional information for the new array after the transformation.

[0124] Based on the new location information X i The detection signal at the new position is calculated using linear interpolation. The interpolation method is as follows:

[0125]

[0126] In the formula, j = [X i ] is X i The integer part, The new magnetic induction detection signal data H is obtained from the magnetic induction detection signal data at the corresponding position in the original data. * The vector length is

[0127] Case 2: Regarding array length Signal data, for The array is concatenated to obtain a new signal:

[0128]

[0129] The signal length becomes N*l, until... Then, following the method in case 1, scale the length to...

[0130] Case 3: Regarding array length The data without abnormal signals is divided from the start and end positions of the array towards the opposite side. The length yields two new data points:

[0131]

[0132] Case 4: Regarding array length The abnormal signal data H is obtained from the array. Starting position Cut from length 1 to obtain length 2. The signal, until the remaining length Then, following the methods in cases 1-3, change the length of the remaining part to...

[0133] Case 5: Regarding array length If there is abnormal signal data H, then the abnormal region set G is expressed as:

[0134] G = {g i ,g i +D i ], i = 1, 2, ..., q}

[0135] In the formula, q represents the number of abnormal regions in H, and g i Let D be the starting position of the i-th anomaly. i Let the location length of the i-th anomaly be denoted by . First, for each anomaly region, divide it outwards from the start and end points of the anomaly signal. Length, if or If an interval completely contains other abnormal signal intervals, then the abnormal intervals are merged and pruned. Specifically: when At that time, abnormal regions are merged into a single signal and pruned; similarly, when... At that time, abnormal regions are merged into a single signal and pruned, resulting in an array length of [length missing]. like or The interval does not completely contain other abnormal signals, from the abnormal signal [g i ,g i+D i Extending from the midpoint of the graph to both sides, we obtain a length of... interval If this interval contains endpoints of other abnormal signals, then that endpoint is used as a clipping point, and the interval is extended to the other side to ensure that the length of the portion containing the abnormal signal is [length missing]. For clipping intervals If the right endpoint falls within another abnormal interval at this point, then the clipping interval is further narrowed to the left until there are no abnormalities, resulting in the final clipping interval [g]. i-1 +D i-1 ,g i+1 For the final trimmed signal array, repeat cases 1-2; for The signal directly The abnormal signal region is trimmed into a signal array, and then the remaining part is processed according to cases 1-5; for the remaining no-signal region after the above trimming, cases 1-4 are executed in a loop.

[0136] In this embodiment, by first adjusting the scale of the magnetic induction detection signal data according to the relationship between the array length of the magnetic induction detection signal data and the average array length of the inspected pipe section, the problem that the horizontal coordinate scale of the magnetic induction detection signal data obtained each time is different due to the different mileage measured each time is also different, which leads to the inability to directly use the obtained magnetic induction detection signal data for intelligent identification of pipeline maintenance level based on neural network can be solved.

[0137] Based on the above embodiments, in this embodiment, sample magnetic induction detection signal data with pre-determined abnormal area locations and maintenance levels are used as input, and the abnormal area locations and maintenance levels in the sample magnetic induction detection signal data are used as output. A neural network model is trained and tested based on an intelligent learning algorithm to obtain an integrity intelligent evaluation model, including:

[0138] Based on the statistical results of historically acquired magnetic induction detection signal data, the ratio of the number of magnetic induction detection signal data containing defects to the number of magnetic induction detection signal data not containing defects is determined.

[0139] Based on the aforementioned quantity ratio, the sample magnetic induction detection signals of the pipeline anomaly area location and maintenance level are randomly divided into training set and test set to train and test the integrity intelligent evaluation model.

[0140] In this embodiment, based on the statistical results of historically acquired magnetic induction detection signal data, the ratio of magnetic induction detection signal data containing defects to those without defects is determined. Then, according to the ratio, the sample magnetic induction detection signals with determined pipeline abnormality area locations and maintenance levels are randomly divided into training and testing sets. The training set is used to train a neural network model, and the testing set is used to test the effectiveness of the trained integrity intelligent evaluation model. By randomly dividing the samples with determined pipeline abnormality area locations and maintenance levels into training and testing sets according to the ratio, the effectiveness of model training and subsequent testing can be improved, enabling both model training and model testing to achieve the expected results.

[0141] Based on the above embodiments, in this embodiment, during the training of the neural network model, the stochastic gradient descent (SGD) method is used to update the network parameters and train the neural network model; wherein, in the selection of the anchor, an anchor of a certain size with an aspect ratio of 1:2 is selected according to the characteristics of the magnetic induction detection signal.

[0142] Based on the above embodiments, in this embodiment, before inputting the magnetic induction detection signal data into the integrity intelligent evaluation model, the method further includes:

[0143] When applying the model, for the magnetic induction detection signal data, the array length ratio of the magnetic induction detection signal data is first standardized.

[0144] If the signal length Then, scale the signal array to the standard size using the methods described in cases 1 and 2 above;

[0145] If the signal length Then, starting from the left endpoint of the signal array, establish a length of... The sliding window, each The signal is trimmed once by length until the right side of the sliding window reaches the boundary of the signal array, resulting in... A number of signal arrays are input into a network for prediction. Based on the network prediction results, the abnormal signal location and the maintenance level of the abnormal signal are output.

[0146] In this embodiment, when applying the network model, for a segment of magnetic induction signal, if the signal length... Then, according to cases 1 and 2 described above, the signal array is scaled to the standard size; if the signal length... To ensure that potentially faulty areas are completely detected by the network at least once and to reduce data signal distortion, thereby improving network accuracy, scaling is no longer performed. Instead, a length of [length missing] is established starting from the left endpoint of the signal array. The sliding window, each The signal is trimmed once by length until the right side of the sliding window reaches the boundary of the signal array, resulting in... (Round down) signal arrays. Input these signal arrays into the network for prediction, and output the range of abnormal signal locations in the data and the maintenance level of the abnormal signal based on the network prediction results.

[0147] Based on the above embodiments, in this embodiment, the intelligent pipeline integrity evaluation method based on detection signals further includes: a process of adjusting and migrating the intelligent integrity evaluation model;

[0148] The process of adjusting and migrating the integrity intelligent evaluation model includes:

[0149] After further accumulation of signal data and pipeline excavation data, pipeline influencing factors such as metal, trees, and pipeline properties can be vectorized using One-Hot encoding or continuous numerical encoding. This involves statistically analyzing factors that may affect magnetic induction, such as trees, power lines, and pipeline properties around the pipeline during excavation, recording the data for these factors, and obtaining new feature vectors. For example, pipeline property data (pipeline material, diameter, wall thickness) can be categorized into One-Hot and continuous encodings, as shown below:

[0150]

[0151] Where m represents the pipe type, n represents the pipe diameter, and r represents the different pipe thicknesses. The feature maps of specific signal regions selected by the RPN network are concatenated with the feature vectors of surrounding environmental factors. The concatenated feature vectors are then processed by a fully connected neural network layer, and finally, the probability L of a signal sample belonging to each maintenance level is calculated using the Softmax function. i Similarly, the network is trained using the SGD method, and the network outputs L. i In the results, the maximum L i The location represents the most likely maintenance level for that anomaly. The probability L of a signal sample belonging to each maintenance level is... i The calculation is as follows:

[0152]

[0153] In the formula, p i It is the output of the fully connected network layer. N represents the number of maintenance levels, such as 3 levels: immediate repair, planned repair, and monitoring.

[0154] In this embodiment, this processing makes the maintenance level predicted by the integrity intelligent evaluation model more accurate and more meaningful. This effectively solves the problem that the accuracy and efficiency of manual inspection of interference signals are low due to large differences in personnel skill levels and huge workload. Misjudgment and omission of signals not only cause enterprises to carry out blind maintenance, but also cause economic losses and social impact due to work stoppages and production stoppages.

[0155] Based on the above embodiments, in this embodiment, the intelligent evaluation method for pipeline integrity based on detection signals further includes: optimizing the intelligent integrity evaluation model so that the intelligent integrity evaluation model outputs defect types;

[0156] With further data accumulation, the network's output can be further refined to more precise defect types, such as dents, metal corrosion, and welds. A multi-label Sigmoid loss function is added after the fully connected layer of the network to obtain the probability of predicting each type of defect based on the sample magnetic induction detection signal data.

[0157]

[0158] in, Let M be the column vector output by the last fully connected layer of the network, where M is the number of defect types.

[0159] The network is trained using stochastic gradient descent (SGD) with cross-entropy loss function, and the network output is... The defect type corresponding to all positions in the vector that are greater than the threshold;

[0160] After completing the above processing, the network outputs the location of the abnormal area, the type of defect, and the corresponding maintenance level during prediction.

[0161] In this embodiment, after further data accumulation, the network output can be shifted towards more precise defect types, such as dents, metal corrosion, and weld defects. Through the above-mentioned processing provided in this embodiment, more precise defect type information can be output when using the model for prediction, thereby providing more effective guidance for subsequent maintenance.

[0162] It is understood that the intelligent pipeline integrity evaluation method based on detection signals provided in this embodiment specifically includes the following processing steps:

[0163] The system collects magnetic induction detection signal data, anomaly locations within the signal data, and maintenance decision levels determined by experts based on experience or excavation. Anomalies in the signal data with maintenance level classifications are used as labels. A neural network model is constructed, and the collected data is standardized and preprocessed to a uniform size. The data is then randomly divided into training and testing sets according to a set ratio. The model is trained using the training set data and tested using the testing set data, outputting intelligent integrity evaluation results. Further, defect data from pipeline excavation, surrounding environmental data, and maintenance decision data derived through empirical formulas are collected. The accumulated actual defect integrity intelligent evaluation results and their corresponding data signals are substituted into the training module for model optimization and further accurate prediction of defect types.

[0164] The following will combine Figure 2 The intelligent pipeline integrity evaluation process based on detection signals provided in the embodiments of the present invention will be described in detail as follows:

[0165] The magnetic induction detection signal data is treated as a two-dimensional array. This array is vectorized, and features are extracted from the vector using a grayscale VGG convolutional neural network to obtain a feature map vector. After being extracted by the convolutional head, the feature map is fed into the RPN (Region Proposal Network) to obtain the target region, which is then used for prediction.

[0166] Based on the location of the pipeline anomaly area and the maintenance level of the anomaly signal (divided into three categories: immediate repair, planned repair, and monitoring use) determined by the signal analysis experts of the testing company through experience or excavation, the loss function of the samples in the network is calculated. By statistically analyzing the ratio of abnormal to normal signals, the existing data is randomly divided into training and test sets according to this ratio. The network model is trained using the training set. The input of the model is the normalized signal data, and the output is the start and end positions of the defect location on the original signal data (i.e., the number of columns in the two-dimensional array) and the corresponding maintenance level. To reduce the loss function, SGD (Stochastic Gradient Descent) is used to update the network parameters and train the target detection network. To accelerate network training, only one anchor size with an aspect ratio of 1:2 is selected based on the characteristics of the magnetic induction signal. The test set is used to test the network performance. After further data collection and accumulation, to improve network accuracy, newly added data is used in the training module for optimization to further improve the network's detection accuracy.

[0167] When applying a network model, for a segment of magnetic induction signal, if the signal length... Then, scale the signal array to the standard size using the method described in Part 1; if the signal length... To ensure that potentially faulty areas are completely detected by the network at least once and to reduce data signal distortion, thus improving network accuracy, scaling is no longer performed; instead, a length of [length missing] is established starting from the left endpoint of the signal array. The sliding window, each The signal is trimmed once by length until the right side of the sliding window reaches the boundary of the signal array, resulting in... (Round down) signal arrays. Input these signal arrays into the network for prediction, and output the range of abnormal signal locations in the data and the maintenance level of the abnormal signal based on the network prediction results.

[0168] After further accumulation of signal data and pipeline excavation data, pipeline influencing factors such as metal, trees, and pipeline properties can be vectorized using One-Hot encoding or continuous numerical encoding. This involves statistically analyzing factors that may affect magnetic induction, such as trees, power lines, and pipeline properties around the pipeline during excavation, recording the data for these factors, and obtaining new feature vectors. For example, pipeline property data (pipeline material, diameter, wall thickness) can be categorized into One-Hot and continuous encodings, as shown below:

[0169]

[0170] Where m represents the pipe type, n represents the pipe diameter, and r represents the different pipe thicknesses. The feature maps of specific signal regions selected by the RPN network are concatenated with the feature vectors of surrounding environmental factors. The concatenated feature vectors are then processed by a fully connected neural network layer, and finally, the probability L of a signal sample belonging to each maintenance level is calculated using the Softmax function. i Similarly, the network is trained using the SGD method, and the network outputs L. i In the results, the maximum L i The location represents the most likely maintenance level for that anomaly. The probability L of a signal sample belonging to each maintenance level is... i The calculation is as follows:

[0171]

[0172] In the formula, p i It is the output of the fully connected network layer. N represents the number of maintenance levels, such as 3 levels: immediate repair, planned repair, and monitoring.

[0173] With further data accumulation, the network's output can be further refined to more precise defect types, such as dents, metal corrosion, and welds. A multi-label Sigmoid loss function is added after the fully connected layer of the network to obtain the probability that the signal sample is predicted as each type of defect.

[0174]

[0175] in, This is the column vector output by the last fully connected layer of the network, where M is the number of anomaly types.

[0176]

[0177] The network is trained using the cross-entropy loss function and the SGD method, and the network output is... The defect type corresponding to all positions in the vector that are greater than a threshold (usually 0.5).

[0178] After completing the above steps, the network can output the defect type and corresponding maintenance level during prediction, such as "weld defect, for monitoring purposes".

[0179] Therefore, this embodiment addresses the technical shortcomings of existing methods by providing an intelligent pipeline integrity evaluation method based on detection signals. This embodiment standardizes and preprocesses historically accumulated magnetic induction detection data signals, then trains a target detection convolutional neural network to make decisions based on the processing results, incorporating surrounding environmental factors during the decision-making process to continuously improve recognition accuracy. Specifically, the technical solution adopted in this embodiment includes: collecting magnetic induction detection signal data, abnormal locations within the signal data, and maintenance decision levels determined by experts based on experience or excavation at these abnormal locations; using abnormal points in the maintenance level-classified signal data as labels; constructing a neural network model, standardizing and preprocessing the collected data to a uniform size, then randomly dividing the data into training and test sets proportionally, training the model using the training set data, testing the model using the test set data, and outputting intelligent integrity evaluation results; further collecting pipeline excavation defect data, surrounding environmental data, and maintenance decision data obtained through empirical formulas, and substituting the accumulated actual defect integrity evaluation results and their corresponding data signals into the training module for model optimization and further accurate prediction of defect types.

[0180] Therefore, this embodiment proposes a highly adaptable intelligent pipeline integrity evaluation method based on magnetic induction detection signal data, free from the influence of subjective human factors. This method improves the accuracy of detection result analysis, avoids blind maintenance, greatly enhances economy and work efficiency, reduces labor costs, and ultimately improves pipeline safety. This embodiment preprocesses historically accumulated magnetic induction detection signal data, then trains a target detection convolutional neural network to make decisions based on the processing results, incorporating surrounding environmental factors during the decision-making process to continuously improve recognition accuracy.

[0181] Figure 3 A schematic diagram of a pipeline integrity intelligent evaluation device based on detection signals according to an embodiment of the present invention is shown, as follows: Figure 3As shown, the intelligent pipeline integrity evaluation device based on detection signals provided in this embodiment of the invention includes:

[0182] The acquisition module 201 is used to acquire the magnetic induction detection signal data of the pipeline;

[0183] The intelligent evaluation module 202 is used to input the magnetic induction detection signal data into the integrity intelligent evaluation model to obtain the integrity intelligent evaluation result of the pipeline; wherein, the integrity intelligent evaluation result includes the location of abnormal areas and the maintenance level;

[0184] The integrity intelligent evaluation model is obtained by training and testing a neural network model, using sample magnetic induction detection signal data as input and the location and maintenance level of the abnormal area corresponding to the sample magnetic induction detection signal data as output.

[0185] Since the intelligent pipeline integrity evaluation device based on detection signals provided in this embodiment can be used to execute the intelligent pipeline integrity evaluation method based on detection signals provided in the above embodiment, its working principle and beneficial effects are similar, and will not be described in detail here.

[0186] Based on the same inventive concept, another embodiment of the present invention provides an electronic device, see [link to previous document]. Figure 4 The electronic device specifically includes the following components: a processor 301, a memory 302, a communication interface 303, and a communication bus 304.

[0187] The processor 301, memory 302, and communication interface 303 communicate with each other through the communication bus 304; the communication interface 303 is used to realize information transmission between the devices.

[0188] The processor 301 is used to call the computer program in the memory 302. When the processor executes the computer program, it implements all the steps of the above-mentioned intelligent evaluation method for pipeline integrity based on detection signals. For example, when the processor executes the computer program, it implements the following steps: acquiring magnetic induction detection signal data of the pipeline; inputting the magnetic induction detection signal data into the integrity intelligent evaluation model to obtain the pipeline integrity intelligent evaluation result; wherein, the integrity intelligent evaluation result includes the location of abnormal areas and the maintenance level; wherein, the integrity intelligent evaluation model is obtained by training and testing a neural network model by taking sample magnetic induction detection signal data as the input of the model and taking the location of abnormal areas and the maintenance level corresponding to the sample magnetic induction detection signal data as the output of the model.

[0189] Based on the same inventive concept, another embodiment of the present invention provides a non-transitory computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements all the steps of the above-described intelligent pipeline integrity evaluation method based on detection signals. For example, when the processor executes the computer program, it implements the following steps: acquiring magnetic induction detection signal data of the pipeline; inputting the magnetic induction detection signal data into an integrity intelligent evaluation model to obtain an integrity intelligent evaluation result of the pipeline; wherein the integrity intelligent evaluation result includes the location of abnormal areas and the maintenance level; wherein the integrity intelligent evaluation model is obtained by training and testing a neural network model, using sample magnetic induction detection signal data as input and the location of abnormal areas and the maintenance level corresponding to the sample magnetic induction detection signal data as output.

[0190] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0191] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the embodiments of the present invention according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0192] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the pipeline integrity intelligent evaluation method based on detection signals described in various embodiments or some parts of embodiments.

[0193] Furthermore, in this invention, terms such as "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0194] Furthermore, in this invention, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0195] Furthermore, in the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0196] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent evaluation of pipeline integrity based on detection signals, characterized by, The method comprises: acquiring magnetic induction detection signal data of a pipeline; inputting the magnetic induction detection signal data into an integrity intelligent evaluation model to obtain an integrity intelligent evaluation result of the pipeline; wherein the integrity intelligent evaluation result comprises an abnormal area position and a repair level; wherein the integrity intelligent evaluation model is obtained by taking sample magnetic induction detection signal data as input of the model, taking an abnormal area position and a repair level corresponding to the sample magnetic induction detection signal data as output of the model, and training and testing a neural network model; the method further comprises a training process of the integrity intelligent evaluation model; wherein the training process of the integrity intelligent evaluation model comprises: collecting a preset number of sample magnetic induction detection signal data; wherein the preset number of sample magnetic induction detection signal data needs to cover various abnormal area positions and various repair levels; determining the pipeline abnormal area position and the repair level corresponding to the preset number of sample magnetic induction detection signal data according to detection signal analysis experts' experience or excavation; taking the sample magnetic induction detection signal data as input of the model, taking the abnormal area position and the repair level corresponding to the sample magnetic induction detection signal data as output of the model, and training and testing a neural network model based on an intelligent learning algorithm to obtain an integrity intelligent evaluation model; after collecting the preset number of sample magnetic induction detection signal data, the method further comprises the following steps: performing length unification processing on the sample magnetic induction detection signal data, specifically comprising: the magnetic induction detection signal of each pipeline is usually a two-dimensional array, in the form of: wherein m is the number of sampling times, and the magnetic induction detection signal obtained at each sampling point is: wherein n is the dimension of the magnetic induction detection signal, h ij represents the numerical size of the jth signal feature of the magnetic induction detection signal; the sampling frequency of the magnetic induction detection signal data is unified, and the length of H represents the length of the detected pipeline section. The magnetic induction detection signal data is standardized to make the length of each detection signal vector the same for network training.

2. The method of claim 1, wherein, First, the average length l of all detected pipeline magnetic induction detection signal data is counted and rounded to an integer, and then the following cases 1 to 5 are executed: Case 1: for signal data of array length , the array length is scaled to using interpolation. Specifically, first the position of the sample point of the new array in the original value coordinate is obtained: In the formula, X i is the position information set of the new array after transformation; According to the new position information X i The detection signal at the new position is calculated using linear interpolation The interpolation method is where j = [X i ] is the integer part of X i is the magnetic induction detection signal data at the corresponding position in the original data, and the vector length of the obtained new magnetic induction detection signal data H * ​​ Case 2: Concatenate the signal data of array length to get a new signal: array: The signal length becomes N*l, until Again, the method in case 1 is used to scale the length to Case 3: Regarding array length The data without abnormal signals is divided into two new data points by dividing the array from the beginning and end positions towards the opposite side: Case 4: Regarding array length The data H without abnormal signals is obtained from the array. Starting position Cut from length 1 to get length 2. The signal, until the remaining length Then, following the methods in cases 1-3, change the length of the remaining part to... Case 5: For the array length of the exceptional signal data H, the exceptional region set G is expressed as G = {[g i ,g i + D i ], i = 1, 2,..., q} where q is the number of abnormal regions in H, g i is the start position of the i-th abnormality, D i is the length of the i-th abnormality, first, for each abnormal region, respectively, from the start and end points of the abnormal signal to the two sides of the division length, if or interval contains other signal abnormal interval, merge the abnormal interval and clip, specifically: when , the abnormal region is merged into the same signal and trimmed, similarly when , the abnormal region is merged into the same signal and trimmed, the length of the trimmed array is if or interval does not completely contain other abnormal signals, respectively, from the midpoint of the abnormal signal[g i ,g i +D i ] to the two sides, get the interval length if the interval contains the end point of other abnormal signals, the end point is taken as a trimming point, and the other side is extended for trimming, ensuring that the length of the signal abnormality containing part is for the trimmed interval if the right end point falls within other abnormal intervals at this time, it is further shrunk to the left without abnormality, to get the final trimmed interval[g i-1 +D i-1 ,g i+1 ], repeat situation 1-situation 2 for the final trimmed signal array; for signal, directly clip the signal abnormal region of signal into a signal array, and then perform situation 1-situation 5 processing on the remaining part; for the remaining signal region after the above clipping, situation 1-situation 4 is executed in a loop.

3. The method of claim 1, wherein, taking the sample magnetic induction detection signal data as input of the model, taking the abnormal area position and the repair level corresponding to the sample magnetic induction detection signal data as output of the model, and training and testing a neural network model based on an intelligent learning algorithm to obtain an integrity intelligent evaluation model, comprising: determining the number proportion of magnetic induction detection signal data containing defects and magnetic induction detection signal data not containing defects according to the statistical results of the historical magnetic induction detection signal data acquired; according to the number proportion, randomly dividing the sample magnetic induction detection signal data for which the pipeline abnormal area position and the repair level have been determined into a training set and a test set to train and test the integrity intelligent evaluation model.

4. The method of claim 1 to 3, wherein, Before inputting the magnetic induction detection signal data into the integrity intelligent evaluation model, the method further comprises: when applying the model, for the magnetic induction detection signal data, first perform array length proportion unification on the magnetic induction detection signal data; if the signal length then the signal array is scaled to the standard size according to the method of case 1 and case 2 as described in claim 2; If the signal length then from the left end point of the signal array, a sliding window with length is established, and the signal is cut once every length until the right side of the sliding window reaches the boundary of the signal array, obtaining signal arrays. These signal arrays are input into the network for prediction, and the interval of the abnormal signal position on the data and the repair level of the abnormal signal are output according to the network prediction result.

5. The method of claim 1 to 3, wherein, further comprising: a processing process of adjusting and migrating the integrity intelligent evaluation model; The processing procedure of adjusting and migrating the integrity intelligent evaluation model comprises: After accumulating pipeline excavation data, the surrounding environmental factors affecting magnetic induction, including metal, trees and pipeline properties, are One-Hot coded or continuously coded to obtain a first feature vector; the feature vector corresponding to the target region selected by the region generation network RPN on the feature map obtained by the convolutional neural network is spliced with the first feature vector to obtain a second feature vector; the second feature vector is calculated through a fully connected neural network layer, and then through a Softmax function, to obtain the probability L of the sample magnetic induction detection signal data belonging to each maintenance level i , the fully connected network is trained using the stochastic gradient descent method SGD, and the network output L i In the result, the maximum L i is the most likely maintenance level of the anomaly at this location. where L is the probability that a signal sample belongs to each repair level i The calculation is: where p i is the output of the fully connected network layer, and N represents the number of repair levels.

6. The method of claim 5, wherein, Further comprising: The integrity intelligent evaluation model is optimized so that the integrity intelligent evaluation model outputs defect types; A multi-label Sigmod loss function is added after a full connection layer of the network, to obtain a probability of predicting each type of defect from sample magnetic induction detection signal data wherein, is a column vector output of the last fully connected layer of the network, M is the number of defect types, The network is trained by using a cross-entropy loss function and a stochastic gradient descent method (SGD), and the network output is The defect types corresponding to all positions greater than the threshold value in the vector; After the above processing, the network outputs abnormal area positions, defect types and corresponding repair levels during prediction.

7. A pipeline integrity intelligent evaluation device based on a detection signal, characterized by, Comprise: The acquisition module is configured to acquire magnetic induction detection signal data of the pipeline; The intelligent evaluation module is configured to input the magnetic induction detection signal data into an integrity intelligent evaluation model to obtain an integrity intelligent evaluation result of the pipeline; wherein the integrity intelligent evaluation result comprises an abnormal area position and a repair level; The integrity intelligent evaluation model is obtained by training and testing a neural network model using sample magnetic induction detection signal data as input and corresponding abnormal area positions and repair levels as output; Further comprising a training module configured to perform a training procedure on the integrity intelligent evaluation model; The training procedure of the integrity intelligent evaluation model comprises: A predetermined number of sample magnetic induction detection signal data are collected; wherein the predetermined number of sample magnetic induction detection signal data need to cover various abnormal area positions and various repair levels; According to the detection signal analysis experts' experience or excavation, the pipeline abnormal area positions and repair levels corresponding to the predetermined number of sample magnetic induction detection signal data are determined; Sample magnetic induction detection signal data are used as input, and abnormal area positions and repair levels corresponding to the sample magnetic induction detection signal data are used as output, and a neural network model is trained and tested based on an intelligent learning algorithm to obtain an integrity intelligent evaluation model; After collecting the predetermined number of sample magnetic induction detection signal data, the training module is further configured to perform the following steps: The sample magnetic induction detection signal data are subjected to length uniformity processing, specifically comprising: The magnetic induction detection signal of each pipeline is usually a two-dimensional array, in the form of: Wherein, m is the sampling times, and the magnetic induction detection signal obtained at each sampling point is: wherein n is the dimension of the magnetic induction detection signal, h ij denotes the numerical value of the jth signal feature of the magnetic induction detection signal; The sampling frequency of the magnetic induction detection signal data is uniform, and the length of H represents the length of the detected pipeline section. The magnetic induction detection signal data are subjected to standardization processing so that the length of each detection signal vector is the same, facilitating network training.

8. The apparatus for intelligent evaluation of pipeline integrity based on detection signals according to claim 7, characterized by, First, the average length of the detected pipe section magnetic induction detection signal data is counted Rounded to an integer, and then the following cases 1 to case 5 are executed: Case 1: for signal data of array length , the array length is scaled to in an interpolated manner. Specifically, first the position of the sample point of the new array in the original value coordinate is obtained: In the formula, X i is the position information set of the new array after transformation; According to the new position information X i The detection signal at the new position is calculated using linear interpolation The interpolation method is where j = [X i ] is the integer part of X i is the magnetic induction detection signal data at the corresponding position in the original data, and the vector length of the obtained new magnetic induction detection signal data H * is ​ Case 2: Concatenate the signal data of array length to get a new signal: array: The signal length becomes N*l, until Again, the method in case 1 is used to scale the length to Case 3: For the array length of the non-exception signal data, the start position and the end position of the array are respectively divided to the opposite side length, and two new data are obtained, namely: Case 4: For the array length of the non-exception signal data H, from the array start position, cut by length, get the signal of length , until the length of the remaining part Also change the length of the remaining part to Case 5: For the array length of the exceptional signal data H, the exceptional region set G is expressed as G = {[g i ,g i + D i ], i = 1, 2,..., q} where q is the number of abnormal regions in H, g i is the start position of the i-th abnormality, D i is the length of the i-th abnormality, first, for each abnormal region, respectively, from the start and end points of the abnormal signal to the two sides of the division length, if or interval contains other signal abnormal interval, merge the abnormal interval and clip, specific: when , the abnormal region is merged into the same signal and cut, and similarly when , the abnormal region is merged into the same signal and cut, the length of the array after cutting is if or interval does not completely contain other abnormal signals, respectively, from the midpoint of the abnormal signal [g i ,g i +D i ] to the two sides, get the interval length if the interval contains the end point of other abnormal signals, the end point is taken as a clipping point, and the other side is extended for clipping, to ensure that the length of the signal abnormal part is for the clipping interval if the right end point falls in other abnormal interval at this time, it is further shrunk to the abnormal place on the left to get the final clipping interval [g i-1 +D i-1 ,g i+1 ], repeat the case 1-case 2 for the final clipped signal array; for signal, directly clip the signal abnormal region of signal to a signal array, and then perform case 1-case 5 processing on the remaining part; for the remaining signal region after the above clipping, loop case 1-case 4.

Images