A method, system and device for early detection of metal pipeline corrosion

Abstract

The present invention discloses a method, system, and device for early detection of metal pipeline corrosion, relating to the field of passive wireless sensing technology. The method comprises: receiving a to-be-processed microwave signal corresponding to each initial microwave signal; preprocessing each to-be-processed microwave signal to obtain a microwave signal feature set; the microwave signal feature set includes microwave resonance points, the correlation coefficient and Euclidean distance between each microwave resonance point and a healthy resonance point, and the mean, median, and variance of each microwave resonance point; inputting multiple microwave signal feature sets into a pipeline corrosion detection model to determine the pipeline status of the metal pipeline; the pipeline status includes water inflow and no water inflow; the pipeline corrosion detection model is determined by training a random forest model based on a sample set. The present invention achieves intelligent and automated prediction of metal pipeline corrosion.

Description

Technical Field

[0001] The present invention relates to the field of passive wireless sensing technology, and in particular to a method, system and device for early detection of metal pipeline corrosion. Background Art

[0002] Metal pipes play an important role in the energy, chemical, and electric power industries. In actual applications, metals are easily corroded by soil, rain, snow, temperature and humidity changes, physical impact, etc. If not discovered in time, it can easily lead to pipeline leakage, resulting in serious consequences and huge economic losses.

[0003] Water and oxygen are the primary causes of pipeline corrosion, so monitoring the health of metal surface anti-corrosion coatings is a means of achieving early detection of metal corrosion. Microwave-based early detection of metal pipeline corrosion offers advantages such as simple structure, low cost, easy installation, and multi-point detection of coating defects and early corrosion on the pipeline surface. These methods can be deployed on a large scale as nodes in a structural health monitoring network. However, traditional structural health monitoring methods, which rely on manual processing and judgment of sensor data, suffer from low efficiency and limited reliability. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, system and device for early detection of metal pipeline corrosion, so as to realize intelligent and automatic prediction of metal pipeline corrosion.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] In a first aspect, the present invention provides a method for early detection of metal pipeline corrosion, comprising:

[0007] receiving a microwave signal to be processed corresponding to each initial microwave signal; the microwave signal to be processed is a signal of the initial microwave signal reflected by the metal pipe;

[0008] Preprocessing is performed on each of the microwave signals to be processed to obtain a microwave signal feature set; the microwave signal feature set includes a microwave resonance point, a correlation coefficient and a Euclidean distance between each microwave resonance point and a healthy state resonance point, and a mean, a median, and a variance of each microwave resonance point; the healthy state resonance point is a microwave resonance point corresponding to the microwave signal to be processed obtained when no water enters the metal pipe;

[0009] Multiple microwave signal feature sets are input into a pipeline corrosion detection model to determine the pipeline status of the metal pipeline; the pipeline status includes water entering the metal pipeline and no water entering the metal pipeline; the pipeline corrosion detection model is determined by training a random forest model based on a sample set; the sample set includes multiple samples, each of which includes a microwave sample feature set corresponding to any position of the metal pipeline and the pipeline status.

[0010] Optionally, the process of determining the sample set specifically includes:

[0011] determining a plurality of locations to be inspected on the metal pipe;

[0012] For each of the positions to be detected, obtaining a microwave signal feature set of the position to be detected;

[0013] According to the pipeline status of the position to be detected, a 0-1 value of the position to be detected is determined; a value of 0 in the 0-1 value indicates that water has not entered the position to be detected, and a value of 1 in the 0-1 value indicates that water has entered the position to be detected; the microwave signal feature set of multiple positions to be detected on the metal pipe and the 0-1 value of each position to be detected constitute a multi-label sample; multiple multi-label samples constitute a sample set.

[0014] Optionally, the process of determining the sample set specifically includes:

[0015] determining a plurality of locations to be inspected on the metal pipe;

[0016] For each of the positions to be detected, obtaining a microwave signal feature set of the position to be detected;

[0017] According to the pipeline status of multiple locations to be detected, the number of water inlet locations of the metal pipe is determined; the microwave signal feature set of multiple locations to be detected on the metal pipe and the number of water inlet locations of the metal pipe constitute a single-label multi-classification sample; multiple single-label multi-classification samples constitute a sample set.

[0018] Optionally, the pipeline corrosion detection model training process specifically includes:

[0019] Randomly draw N samples from a sample set of size N with replacement;

[0020] According to the set extraction ratio, m feature dimensions are randomly extracted from the M-dimensional features; the N samples and the m feature dimensions constitute a subsample set; the subsample set is used to train the sub-decision tree; the feature dimension refers to the number of microwave signal features in the microwave signal feature set; m<<M;

[0021] Repeat the sampling and feature dimension extraction T times to obtain T sub-decision trees and the classification results corresponding to each sub-decision tree; the T sub-decision trees constitute a preliminary random forest;

[0022] If the sample is a single-label multi-classification sample, when the accuracy of the preliminary random forest reaches a first set threshold range, the preliminary random forest is a pipeline corrosion detection model;

[0023] If the sample is a multi-label sample, when the accuracy of the preliminary random forest reaches a second set threshold range and the Hamming loss reaches a third set threshold range, the preliminary random forest is a pipeline corrosion detection model.

[0024] Optionally, preprocessing each of the microwave signals to be processed to obtain a microwave signal feature set specifically includes:

[0025] Performing noise reduction processing on the microwave signal to be processed by using a sliding average method;

[0026] Feature extraction is performed on the microwave signal to be processed after noise reduction processing to obtain a microwave signal feature set.

[0027] Optionally, the process of calculating the correlation coefficient between the microwave resonance point and the healthy state resonance point includes:

[0028] According to the formula

[0029]

[0030] Calculate the correlation coefficient between the microwave resonance point and the healthy state resonance point;

[0031] Wherein, X = {x1, x2, ..., xn} is the set of microwave resonance points, H0 = {h1, h2, ..., hn} is the set of healthy state resonance points, Cov(X, H0) is the covariance between the microwave resonance point and the healthy state resonance point, Var(X) is the variance of the set of microwave resonance points, and Var(H0) is the variance of the set of healthy state resonance points.

[0032] In a second aspect, the present invention provides a system for early detection of metal pipeline corrosion, comprising:

[0033] a microwave signal acquisition module, configured to receive a microwave signal to be processed corresponding to each initial microwave signal; the microwave signal to be processed is a signal of the initial microwave signal reflected by the metal pipe;

[0034] a feature extraction module configured to preprocess each of the microwave signals to be processed to obtain a microwave signal feature set; the microwave signal feature set includes a microwave resonance point, a correlation coefficient and a Euclidean distance between each microwave resonance point and a healthy state resonance point, and a mean, median, and variance of each microwave resonance point; the healthy state resonance point is a microwave resonance point corresponding to the microwave signal to be processed obtained when no water enters the metal pipe;

[0035] The pipeline corrosion detection module is used to input the multiple microwave signal feature sets into the pipeline corrosion detection model to determine the pipeline status of the metal pipeline; the pipeline status includes water inflow and no water inflow; the pipeline corrosion detection model is determined by training a random forest model based on a sample set; the sample set includes multiple samples, each of which includes a microwave sample feature set corresponding to any position of the metal pipeline and the pipeline status.

[0036] In a third aspect, the present invention provides a device for early detection of corrosion in metal pipelines, comprising an underlying dielectric substrate, a resonator array, a first broadband microstrip antenna, and a second broadband microstrip antenna;

[0037] The resonator array, the first broadband microstrip antenna, and the second broadband microstrip antenna are all arranged on the bottom dielectric substrate; the first broadband microstrip antenna and the second broadband microstrip antenna are placed orthogonally, and the first broadband microstrip antenna and the second broadband microstrip antenna are connected through a first metal layer; the resonator array is arranged on the first metal layer;

[0038] The first broadband microstrip antenna is used to transmit a plurality of initial microwave signals to the metal pipe;

[0039] The second broadband microstrip antenna is used to receive a microwave signal to be processed corresponding to each of the initial microwave signals; the microwave signal to be processed is a signal of the initial microwave signal reflected by the metal pipe.

[0040] Optionally, the first broadband microstrip antenna includes a first layer of metal patch and a first layer of dielectric substrate; the first layer of dielectric substrate is provided on the first layer of metal patch; an elliptical groove is provided on the first layer of metal patch; and a second metal layer is provided on the first layer of dielectric substrate;

[0041] The second broadband microstrip antenna includes a second metal patch and a second dielectric substrate; the second dielectric substrate is arranged on the second metal patch; the second metal patch is provided with an elliptical groove; and the second dielectric substrate is provided with a third metal layer.

[0042] Optionally, the device further comprises a protective coating;

[0043] The protective coating is provided on the surface of the metal pipe;

[0044] The bottom dielectric substrate is arranged on the protective coating, and a hose is embedded in the protective coating; the hose is used to circulate water to simulate different pipeline states of the metal pipeline; the pipeline states include water flowing into the metal pipeline and no water flowing into the metal pipeline.

[0045] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0046] The present invention provides a method, system, and device for early detection of metal pipeline corrosion. The method transmits multiple initial microwave signals to the metal pipeline, then receives the microwave signals to be processed that are reflected by the metal pipeline, and pre-processes the microwave signals to be processed to obtain a microwave signal feature set. The microwave signal feature set includes the microwave resonance point, the mean, median, variance, correlation coefficient, and Euclidean distance between each microwave resonance point and the healthy state resonance point. The health state of the metal pipeline can be determined by the above-mentioned features related to the microwave resonance point. Finally, the microwave signal feature set is input into a trained pipeline corrosion detection model to obtain the pipeline state of the metal pipeline, that is, whether water has entered the metal pipeline or not. The present invention obtains the microwave signal feature set through simulation and experiment, and uses the random forest algorithm to train and learn to obtain a pipeline corrosion detection model, so that the pipeline corrosion detection model can intelligently and accurately detect the health state of the metal pipeline. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 Schematic diagram of the process of the method for early detection of metal pipeline corrosion according to the present invention;

[0049] Figure 2 This is a schematic structural diagram of a system for early detection of metal pipeline corrosion according to the present invention;

[0050] Figure 3 A top view of the device for early detection of metal pipeline corrosion according to the present invention;

[0051] Figure 4 It is a side view of the device for early detection of metal pipeline corrosion according to the present invention. Description of the drawings:

[0053] 1-bottom dielectric substrate, 2-first metal layer, 3-resonator array, 4-first metal patch, 5-elliptical groove, 6-first dielectric substrate, 7-second metal layer, 8-metal pipe, 9-protective coating, 10-hose. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0055] In order to make the objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] Example 1

[0057] like Figure 1 As shown, this embodiment provides a method for early detection of metal pipeline corrosion, comprising:

[0058] Step 100: receiving a microwave signal to be processed corresponding to each initial microwave signal; the microwave signal to be processed is a signal of the initial microwave signal reflected by the metal pipe, specifically, the microwave signal to be processed is S 21 Transfer curve.

[0059] Step 200: Preprocess each of the microwave signals to be processed to obtain a microwave signal feature set; the microwave signal feature set includes a microwave resonance point, a correlation coefficient and Euclidean distance between each microwave resonance point and a healthy state resonance point, and a mean, median, and variance of each microwave resonance point; the healthy state resonance point is a microwave resonance point corresponding to the microwave signal to be processed obtained when no water enters the metal pipe.

[0060] Preferably, step 200 includes:

[0061] 1) Using the sliding average method to perform noise reduction on the microwave signal to be processed. Specifically, according to the formula The collected S 21 The parameters of the transmission curve (microwave signal to be processed) are used to filter out noise to achieve a smooth curve. In the formula, x j is the original input data, i.e. the microwave signal to be processed, X l is the lth microwave signal to be processed after noise reduction processing, k is the sliding window size, and the smoothing effect can be adjusted by adjusting k.

[0062] 2) Feature extraction is performed on the noise-reduced microwave signal to obtain a microwave signal feature set. Specifically, each resonance point in a typical microwave signal corresponds to a notch of a different frequency. Therefore, a sliding window is applied to obtain the horizontal coordinate corresponding to the local minimum value to obtain the microwave resonance points corresponding to all resonant frequencies of the noise-reduced microwave signal. The mean, median, and variance of each microwave resonance point are then calculated, along with the correlation coefficient and Euclidean distance between each microwave resonance point and the healthy resonance point, and these are used as features.

[0063] According to the formula

[0064]

[0065] Calculate the correlation coefficient between the microwave resonance point and the healthy state resonance point; X = {x1, x2, ..., xn} is the set of microwave resonance points, H0 = {h1, h2, ..., hn} is the set of healthy state resonance points, the microwave resonance points correspond one-to-one to the healthy state resonance points, Cov(X, H0) is the covariance between the microwave resonance point and the healthy state resonance point, Var(X) is the variance of the set of microwave resonance points, and Var(H0) is the variance of the set of healthy state resonance points.

[0066] According to the formula

[0067]

[0068] Calculate the Euclidean distance between the microwave resonance point and the healthy state resonance point.

[0069] In step 300, the plurality of microwave signal feature sets are input into a pipeline corrosion detection model to determine the pipeline status of the metal pipeline; the pipeline status includes water inflow and water non-inflow; the pipeline corrosion detection model is determined by training a random forest model based on a sample set; the sample set includes a plurality of samples, each of which includes a microwave sample feature set corresponding to any position of the metal pipeline and the pipeline status.

[0070] In a specific embodiment of the present invention, two classification methods are set up as needed to achieve the prediction of the health of the metal pipe coating, namely multi-label classification and single-label multi-classification. The former multi-label classification is used to distinguish where the water is entering the metal pipe. The Q groups of resonant sensors contained in the sensor array correspond to Q labels respectively. Each label distinguishes whether water has entered under each group of resonant sensors, which can be divided into two categories: water inflow or no water inflow (healthy); the latter single-label multi-classification predicts the overall water inflow situation of the metal pipe, only judging the number of water inflow points of the metal pipe, without distinguishing where the water enters. If the pipeline is healthy (water inflow at 0 points) is marked as 0, and water inflow at P points is marked as P, it can be divided into P+1 categories. Staff can choose one or two according to their needs for early detection of metal pipe corrosion.

[0071] (1) Multi-label classification

[0072] The process of determining the sample set specifically includes:

[0073] 1) Determine multiple locations to be inspected on the metal pipe.

[0074] 2) For each of the positions to be detected, a microwave signal feature set of the position to be detected is obtained.

[0075] 3) Determine a 0-1 value for the position to be detected based on the pipeline status of the position to be detected; a value of 0 in the 0-1 value indicates that water has not entered the position to be detected, and a value of 1 in the 0-1 value indicates that water has entered the position to be detected; a microwave signal feature set of multiple positions to be detected on the metal pipe and the 0-1 value of each position to be detected constitute a multi-label sample; and multiple multi-label samples constitute a sample set.

[0076] The training process of the pipeline corrosion detection model corresponding to this sample set specifically includes:

[0077] 1) Randomly draw N samples from a sample set of size N with replacement.

[0078] 2) According to the set extraction ratio, m feature dimensions are randomly extracted from the M-dimensional features; the N samples and the m feature dimensions constitute a subsample set; the subsample set is used to train the sub-decision tree; the feature dimension refers to the number of microwave signal features in the microwave signal feature set; m<<M.

[0079] 3) Repeat the sampling and feature dimension extraction T times to obtain T sub-decision trees and the classification results corresponding to each sub-decision tree; the T sub-decision trees constitute a preliminary random forest.

[0080] 4) When the accuracy of the preliminary random forest reaches the second set threshold range and the Hamming loss reaches the third set threshold range, the preliminary random forest is a pipeline corrosion detection model, and the preliminary random forest model selects the category with the most votes through the majority voting method as the final classification.

[0081] Among them, the accuracy rate is the proportion of samples with completely correct label predictions to the total number of samples. According to the formula

[0082]

[0083] Calculate the subset accuracy of the classification result corresponding to the decision tree. SubsetAccuracy represents the subset accuracy of the classification result corresponding to the decision tree, function f represents the mapping relationship between sample space X→label space Y, I is the indicator function, when f(X i ) and Y i If they are exactly equal, the function returns 1, otherwise it returns 0.

[0084] Hamming loss is used to measure the mismatch between the predicted label and the actual label, that is, the proportion of incorrectly predicted labels. According to the formula

[0085]

[0086] Calculate the Hamming loss of the classification result corresponding to the decision tree. Hamming Loss represents the Hamming loss of the classification result corresponding to the decision tree, and Δ represents the symmetric difference between the two label sets, which is equivalent to the result of the XOR operation of the two label sets.

[0087] (2) Single-label multi-classification

[0088] The process of determining the sample set specifically includes:

[0089] 1) Determine multiple locations to be inspected on the metal pipe.

[0090] 2) For each of the positions to be detected, a microwave signal feature set of the position to be detected is obtained.

[0091] 3) Determine the number of water-influent positions and non-water-influent positions of the metal pipe based on the pipeline status of multiple positions to be detected; the microwave signal feature set of multiple positions to be detected on the metal pipe, the number of water-influent positions and non-water-influent positions of the metal pipe constitute a single-label multi-classification sample; and multiple single-label multi-classification samples constitute a sample set.

[0092] The training process of the pipeline corrosion detection model corresponding to this sample set is similar to the model training process for multi-label classification in (I) above, with the following differences:

[0093] If the sample is a single-label multi-classification sample, then when the accuracy of the preliminary random forest reaches the first set threshold range, the preliminary random forest is a pipeline corrosion detection model; wherein the accuracy is the proportion of the number of samples correctly predicted by the model (TP, TN) to the total number of samples, according to the formula

[0094]

[0095] Calculate the accuracy of the classification result corresponding to the decision tree; Accuracy represents the accuracy of the classification result corresponding to the decision tree, TP represents that the actual sample is a positive example and the model prediction is also a positive example, TN represents that the actual sample is a negative example and the prediction result is also a negative example, FP represents that the actual sample is a negative example and the model prediction is a positive example, and FN represents that the actual sample is a positive example and the prediction result is a negative example.

[0096] It should be noted that in the present invention, whether it is the first set threshold range corresponding to the accuracy of the preliminary random forest, the second set threshold range corresponding to the correctness, or the third set threshold range corresponding to the Hamming loss, the first, second, and third set thresholds do not refer to priority or order. These three are evaluated simultaneously during the training process of the random forest to constrain the preliminary random forest. In addition, different numerical ranges can be assigned to the first set threshold range, the second set threshold range, and the third set threshold range according to actual needs.

[0097] In summary, through simulation and experiments, the present invention obtains the transmission coefficient under different water inflow conditions, creates a data set, extracts all resonance points, and calculates characteristic values ​​such as the mean, median, variance, correlation coefficient and Euclidean distance of all resonance points with the healthy state resonance point. The random forest algorithm is used to train single-label multi-classification and multi-label classification models, respectively, to detect and predict the health status of metal pipelines, thereby realizing automatic prediction of the pipeline health status.

[0098] Example 2

[0099] like Figure 2 As shown, this embodiment provides a system for early detection of metal pipeline corrosion, comprising:

[0100] The microwave signal acquisition module 101 is configured to receive a microwave signal to be processed corresponding to each initial microwave signal; the microwave signal to be processed is a signal of the initial microwave signal reflected by the metal pipe.

[0101] The feature extraction module 201 is used to preprocess each of the microwave signals to be processed to obtain a microwave signal feature set; the microwave signal feature set includes a microwave resonance point, a correlation coefficient and a Euclidean distance between each microwave resonance point and a healthy state resonance point, and a mean, a median, and a variance of each microwave resonance point; the healthy state resonance point is a microwave resonance point corresponding to the microwave signal to be processed obtained when no water enters the metal pipe.

[0102] The pipeline corrosion detection module 301 is used to input the multiple microwave signal feature sets into the pipeline corrosion detection model to determine the pipeline status of the metal pipeline; the pipeline status includes water inflow and no water inflow; the pipeline corrosion detection model is determined by training a random forest model based on a sample set; the sample set includes multiple samples, each of which includes a microwave sample feature set corresponding to any position of the metal pipeline and the pipeline status.

[0103] Example 3

[0104] like Figure 3 and Figure 4 As shown, this embodiment provides a device for early detection of metal pipeline corrosion, comprising a bottom dielectric substrate 1, a resonator array 3, a first broadband microstrip antenna, and a second broadband microstrip antenna. Specifically, the device for early detection of metal pipeline corrosion is a transmission-based microwave sensor. Microwaves are emitted by a transmitting antenna. When they encounter the object being measured, they are absorbed or reflected, causing a change in power. A receiving antenna receives microwaves that have passed through or been reflected by the object being measured, converts them into electrical signals, and processes them in a measurement circuit, thus achieving microwave detection.

[0105] Specifically, the resonator array 3, the first broadband microstrip antenna, and the second broadband microstrip antenna are all disposed on the bottom dielectric substrate 1; the first broadband microstrip antenna and the second broadband microstrip antenna are placed orthogonally, and the first broadband microstrip antenna and the second broadband microstrip antenna are connected via a first metal layer 2; the resonator array 3 is disposed on the first metal layer 2. The resonator array 3 includes multiple groups of resonators, capable of generating multiple stopbands and covering multiple detection areas. In a specific embodiment, the resonator array 3 is composed of 10 groups of mirror-symmetrical spiral resonators, which can generate 10 stopbands in the range of 3.1-3.8 GHz, covering 10 sensing areas.

[0106] The first broadband microstrip antenna is used to transmit multiple initial microwave signals to the metal pipe 8; the second broadband microstrip antenna is used to receive a microwave signal to be processed corresponding to each of the initial microwave signals; the microwave signal to be processed is a signal reflected by the metal pipe 8 from the initial microwave signal. Specifically, the microwave signal to be processed is S 21Transfer curve.

[0107] Preferably, the first broadband microstrip antenna includes a first layer of metal patch 4 and a first layer of dielectric substrate 6; the first layer of dielectric substrate 6 is arranged on the first layer of metal patch 4; an elliptical groove 5 is provided on the first layer of metal patch 4; a second metal layer 7 is provided on the first layer of dielectric substrate 6; the second broadband microstrip antenna includes a second layer of metal patch and a second layer of dielectric substrate; the second layer of dielectric substrate is arranged on the second layer of metal patch; an elliptical groove 5 is provided on the second layer of metal patch; and a third metal layer is provided on the second layer of dielectric substrate.

[0108] In a specific embodiment, the second metal layer 7 and the third metal layer are both rectangular metal layers and are provided on corresponding dielectric substrates by printing, so that the microstrip antenna can operate normally in the range of 3.1-3.8 GHz.

[0109] Preferably, the device also includes a protective coating 9; the protective coating 9 has a multi-layer structure and can achieve excellent corrosion protection in the environment in which the metal pipe is located. Specifically, the protective coating 9 comprises three layers: epoxy powder, adhesive, and extruded polyethylene. The protective coating 9 is applied to the surface of the metal pipe 8; the underlying dielectric substrate 1 is disposed on the protective coating 9, and a hose 10 is embedded in the protective coating 9; the hose 10 is made of plastic and is used to inject and circulate water to simulate different pipe states of the metal pipe; the pipe states include water flowing into the metal pipe and no water flowing into the metal pipe.

[0110] In a specific application, the plastic hoses are placed under 10 groups of resonators. Since the resonance point of the microwave signal obtained after water enters the metal pipe shifts to a lower frequency, detecting water ingress in the coating can achieve early detection of corrosion.

[0111] Furthermore, in this embodiment, each structure is processed using PCB technology. Depending on the operating frequency band, different processing technologies can also be used to implement it, such as wire-cut electric discharge technology or photolithography technology.

[0112] Compared with the prior art, the present invention also has the following advantages:

[0113] (1) The method, system and device for early detection of metal pipeline corrosion of the present invention detect the health status of the pipeline coating by analyzing the offset of the resonant frequency point. The data set is obtained through simulation and experiment, and the random forest algorithm is used to train and learn the data set, thereby realizing intelligent and automatic classification and judgment of the pipeline health status. The single-label classification and multi-label classification adopted can achieve good classification effects, with average classification accuracy rates reaching 91.5% and 96.25% respectively, and the highest can reach 100%, which proves the feasibility of the method.

[0114] (2) The method, system and device for early detection of metal pipeline corrosion of the present invention can realize detection of multiple sensing areas and realize early detection of pipeline corrosion by detecting water ingress into the coating.

[0115] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0116] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for early detection of metal pipeline corrosion, characterized in that: The method comprises: receiving a microwave signal to be processed corresponding to each initial microwave signal; the microwave signal to be processed is a signal of the initial microwave signal reflected by the metal pipe; Preprocessing is performed on each of the microwave signals to be processed to obtain a microwave signal feature set; the microwave signal feature set includes a microwave resonance point, a correlation coefficient and a Euclidean distance between each microwave resonance point and a healthy state resonance point, and a mean, a median, and a variance of each microwave resonance point; the healthy state resonance point is a microwave resonance point corresponding to the microwave signal to be processed obtained when no water enters the metal pipe; Inputting the plurality of microwave signal feature sets into a pipeline corrosion detection model to determine the pipeline status of the metal pipeline; the pipeline status includes water inflow and no water inflow; the pipeline corrosion detection model is determined by training a random forest model based on a sample set; the sample set includes a plurality of samples, each of which includes a microwave sample feature set corresponding to any position of the metal pipeline and the pipeline status; Two classification methods are set up to predict the health of metal pipeline coatings, namely multi-label classification and single-label multi-classification; In terms of multi-label classification, the process of determining the sample set specifically includes: Determine multiple locations to be detected on the metal pipe; for each of the locations to be detected, obtain a microwave signal feature set for the location to be detected; determine a 0-1 value for the location to be detected based on the pipeline state of the location to be detected; a value of 0 in the 0-1 value indicates that water has not entered the location to be detected, and a value of 1 in the 0-1 value indicates that water has entered the location to be detected; the microwave signal feature sets of the multiple locations to be detected on the metal pipe and the 0-1 value of each location to be detected constitute a multi-label sample; and a plurality of the multi-label samples constitute a sample set; In terms of single-label multi-classification, the process of determining the sample set specifically includes: Determine multiple locations to be detected on the metal pipe; for each of the locations to be detected, obtain a microwave signal feature set of the location to be detected; determine the number of water-influent locations and the number of non-water-influent locations of the metal pipe based on the pipeline status of the multiple locations to be detected; the microwave signal feature sets of the multiple locations to be detected on the metal pipe, the number of water-influent locations and the number of non-water-influent locations of the metal pipe constitute single-label multi-classification samples; and multiple single-label multi-classification samples constitute a sample set.

2. The method for early detection of metal pipeline corrosion according to claim 1, characterized in that: The training process of the pipeline corrosion detection model specifically includes: Randomly draw N samples from a sample set of size N with replacement; According to the set extraction ratio, m feature dimensions are randomly extracted from the M-dimensional features; the N samples and the m feature dimensions constitute a subsample set; the subsample set is used to train the sub-decision tree; the feature dimension refers to the number of microwave signal features in the microwave signal feature set; m M; Repeat the sampling and feature dimension extraction T times to obtain T sub-decision trees and the classification results corresponding to each sub-decision tree; the T sub-decision trees constitute a preliminary random forest; If the sample is a single-label multi-classification sample, when the accuracy of the preliminary random forest reaches a first set threshold range, the preliminary random forest is a pipeline corrosion detection model; If the sample is a multi-label sample, when the accuracy of the preliminary random forest reaches a second set threshold range and the Hamming loss reaches a third set threshold range, the preliminary random forest is a pipeline corrosion detection model.

3. The method for early detection of metal pipeline corrosion according to claim 1, characterized in that: The preprocessing of each microwave signal to be processed to obtain a microwave signal feature set specifically includes: Performing noise reduction processing on the microwave signal to be processed by using a sliding average method; Feature extraction is performed on the microwave signal to be processed after noise reduction processing to obtain a microwave signal feature set.

4. The method for early detection of metal pipeline corrosion according to claim 1, characterized in that: The calculation process of the correlation coefficient between the microwave resonance point and the healthy state resonance point includes: According to the formula Calculate the correlation coefficient between the microwave resonance point and the healthy state resonance point; in, X ={x1, x2, ..., xn} is the set of microwave resonance points, H 0 ={h1, h2, ..., hn} is the set of healthy state resonance points, Cov(X,H 0 ) is the covariance between the microwave resonance point and the healthy state resonance point, Var(X) is the variance of the set of microwave resonance points, Var(H 0 ) is the variance of the set of healthy state resonance points.

5. A system for early detection of metal pipeline corrosion, applied to the method for early detection of metal pipeline corrosion according to any one of claims 1 to 4, characterized in that: The system comprises: a microwave signal acquisition module, configured to receive a microwave signal to be processed corresponding to each initial microwave signal; the microwave signal to be processed is a signal of the initial microwave signal reflected by the metal pipe; a feature extraction module configured to preprocess each of the microwave signals to be processed to obtain a microwave signal feature set; the microwave signal feature set includes a microwave resonance point, a correlation coefficient and a Euclidean distance between each microwave resonance point and a healthy state resonance point, and a mean, median, and variance of each microwave resonance point; the healthy state resonance point is a microwave resonance point corresponding to the microwave signal to be processed obtained when no water enters the metal pipe; The pipeline corrosion detection module is used to input the multiple microwave signal feature sets into the pipeline corrosion detection model to determine the pipeline status of the metal pipeline; the pipeline status includes water inflow and no water inflow; the pipeline corrosion detection model is determined by training a random forest model based on a sample set; the sample set includes multiple samples, each of which includes a microwave sample feature set corresponding to any position of the metal pipeline and the pipeline status.

6. The system for early detection of metal pipeline corrosion according to claim 5, characterized in that: The microwave signal acquisition module includes a bottom dielectric substrate, a resonator array, a first broadband microstrip antenna and a second broadband microstrip antenna; The resonator array, the first broadband microstrip antenna, and the second broadband microstrip antenna are all arranged on the bottom dielectric substrate; the first broadband microstrip antenna and the second broadband microstrip antenna are placed orthogonally, and the first broadband microstrip antenna and the second broadband microstrip antenna are connected through a first metal layer; the resonator array is arranged on the first metal layer; The first broadband microstrip antenna is used to transmit a plurality of initial microwave signals to the metal pipe; The second broadband microstrip antenna is used to receive a microwave signal to be processed corresponding to each of the initial microwave signals; the microwave signal to be processed is a signal reflected by the metal pipe from the initial microwave signal; The first broadband microstrip antenna includes a first metal patch and a first dielectric substrate; the first dielectric substrate is disposed on the first metal patch; an elliptical groove is disposed on the first metal patch; and a second metal layer is disposed on the first dielectric substrate; The second broadband microstrip antenna includes a second metal patch and a second dielectric substrate; the second dielectric substrate is arranged on the second metal patch; the second metal patch is provided with an elliptical groove; and the second dielectric substrate is provided with a third metal layer.

7. The system for early detection of metal pipeline corrosion according to claim 6, characterized in that: The microwave signal acquisition module further includes a protective coating; The protective coating is provided on the surface of the metal pipe; The bottom dielectric substrate is arranged on the protective coating, and a hose is embedded in the protective coating; the hose is used to circulate water to simulate different pipeline states of the metal pipeline; the pipeline states include water flowing into the metal pipeline and no water flowing into the metal pipeline.

Images