CN114676730A - Method, device, electronic device and storage medium for identifying bent and deformed pipe segment - Google Patents

Abstract

The present application provides a method, device, electronic device and storage medium for identifying a bending deformation pipe segment, the method includes: acquiring first strain data measured by an inertial measurement unit IMU; obtaining a bending strain curve according to the first strain data; determining the bending strain curve The wave crest with the middle bending strain value greater than the preset value is obtained, and the pipe segment to be identified is obtained. The pipe segment to be identified is the pipe segment corresponding to the preset interval on both sides of the wave crest in the pipeline; the second strain data corresponding to the pipe segment to be identified is input into the pipe segment type identification model, Get the abnormal type of the pipe segment to be identified. In the method of the present application, the pipe section to be identified is determined according to the acquired first strain data, and the second strain data corresponding to the pipe section to be identified is input into the pipe section type identification model, and the abnormal type of the pipe section of the to-be-identified pipe section can be obtained. The type judgment can save time and manpower, and unify the judgment standards, thereby improving the judgment efficiency of abnormal types of pipe sections.

Description

Method, device, electronic device and storage medium for identifying bent and deformed pipe segment

technical field

The present application relates to the technical field of data processing, and in particular, to a method, device, electronic device and storage medium for identifying a bent and deformed pipe segment.

Background technique

Long-distance oil and gas pipelines generally have the characteristics of long-distance transportation, shallow pipeline depth, wide spanning area, complex and changeable geological conditions, etc., which are vulnerable to natural disasters such as earthquakes, collapses, landslides, and debris flows. Geological disasters are one of the main reasons for the failure of buried oil and gas pipelines. The surface shape change and soil displacement caused by geological disasters can cause bending deformation of the pipe body. The pipeline is prone to the most serious deformation after bending deformation under the action of external force There will be stress concentration areas at the points, welds or deformation restraint points, which will lead to the failure of the pipe body in severe cases, resulting in immeasurable economic losses, casualties, and environmental damage. Therefore, in-pipe detection technology is required to monitor whether the pipeline is abnormal, so as to avoid various losses caused by the failure of the pipe body.

In-pipeline detection technology is a common technology that relies on the pressure of the transport medium in the pipeline to drive the operation of the inner detector, so as to measure the defects and abnormalities on the oil and gas pipeline. Inertial measurement unit (Inertial Measurement Unit: IMU) detection technology is a commonly used in-pipe detection technology and an effective method to detect the bending strain of the pipeline. Its core components are three-dimensional orthogonal gyroscopes and accelerometers. After the measurement of the entire pipeline to be measured, the data collected and recorded by the IMU is used for integration and other calculations, and the speed, position and attitude data information of the internal detector at any time can be obtained. Based on the attitude data detected by the IMU, the whole The deformation and strain state of each part of the pipeline in the pipeline to be measured, and the risks caused by pipeline displacement and bending strain are identified and controlled through long-term detection.

The existing identification and analysis methods for IMU strain data use Matlab programming to pre-find all pipe sections with bending strain exceeding 0.125% in the IMU strain data. These pipe sections include elbows, depressions, bending deformation sections and abnormal girth weld sections. , based on the aligned geometric detection data, the mileage positions of the given elbows and depressions are marked, and the strain data within the influence range of geometric feature points such as depressions and elbows is deleted, and the interference of elbows and depressions is excluded. For the remaining unmarked pipe sections with a bending strain value greater than 0.125%, the technicians with practical experience judge the type of pipe sections section by section by drawing pictures, and mark the sections by manual identification, which are suspected elbows, suspected depressions, and rings. Weld abnormal segment and bending deformation segment. The method uses geometric detection and manual identification to identify and classify the IMU data segment by segment to obtain the location of the geological disaster pipeline.

The existing identification and analysis of IMU strain data is still at the stage of manually identifying high-risk pipeline sections of geological hazards one by one, and this method has problems such as long time consumption and low identification accuracy.

SUMMARY OF THE INVENTION

The present application provides a method, device, electronic device and storage medium for identifying a bent and deformed pipe section, which are used to solve the problems of the existing identification and analysis of IMU strain data, which take a long time and have a low identification accuracy.

In a first aspect, the present application provides a method for identifying a bent and deformed pipe segment, including:

First strain data measured by the inertial measurement unit IMU is acquired, where the first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline.

From the first strain data, a bending strain curve is obtained.

Determine the wave crest whose bending strain value is greater than the preset value in the bending strain curve, and obtain the pipe section to be identified, and the pipe section to be identified is the pipe section corresponding to the preset interval on both sides of the wave crest in the pipeline.

The second strain data corresponding to the pipe section to be identified is input into the pipe section type identification model, and the abnormal type of the pipe section of the pipe section to be identified is obtained.

Optionally, the pipe segment type identification model is obtained by training based on a one-dimensional convolutional neural network model.

Optionally, before inputting the second strain data corresponding to the pipe segment to be identified into the pipe segment type identification model, the method includes:

The initial pipe segment type identification model is trained based on the sample database and the labels corresponding to the sample data, and the pipe segment type identification model is obtained; Strain value of the pipe segment.

Optionally, the initial pipe segment type identification model is trained based on the sample database and the labels corresponding to the sample data, and the pipe segment type identification model is obtained, including:

Obtain the sample data corresponding to the four types of anomalies in the pipe section, including sag, elbow, bending deformation and girth weld anomaly in the third strain data, and the sample database includes the sample data corresponding to the four types of pipe section anomalies.

The initial pipe segment type identification model is trained based on the sample data corresponding to the four pipe segment types and the labels corresponding to the sample data, and the pipe segment type identification model is obtained; among which, the labels corresponding to the sample data include 4 types of concave, elbow, bending deformation and abnormal girth weld Pipe segment exception type.

Optionally, the initial pipe segment type identification model is trained based on the sample data corresponding to the four pipe segment types and the labels corresponding to the sample data, and the pipe segment type identification model is obtained, including:

The sample data is input into the initial pipe segment type identification model, and after the feature processing of the input layer, convolution layer, pooling layer, flat layer, fully connected layer and output layer, the predicted type of the pipe segment is obtained.

According to the predicted type and label of the pipe segment, the network parameters of the initial pipe segment type identification model are updated to obtain the pipe segment type identification model.

Optionally, according to the predicted type and label of the pipe segment, update the network parameters of the initial pipe segment type identification model to obtain the pipe segment type identification model, including:

According to the prediction type and label of the pipe segment, the loss function corresponding to the sample data is constructed.

According to the loss function corresponding to the sample data, the network parameters of the initial pipe segment type identification model are updated to obtain the pipe segment type identification model.

In a second aspect, the present application provides a device for identifying bent and deformed pipe segments, including:

The acquiring module is configured to acquire first strain data measured by the inertial measurement unit IMU, where the first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline.

The obtaining module is further configured to obtain a bending strain curve according to the first strain data.

The determining module is used for determining the wave crest whose bending strain value is greater than the preset value in the bending strain curve, and obtains the pipe section to be identified, and the pipe section to be identified is the pipe section corresponding to the preset interval on both sides of the wave crest in the pipeline.

The identification module is used for inputting the second strain data corresponding to the pipe section to be identified into the pipe section type identification model to obtain the abnormal type of the pipe section of the pipe section to be identified.

Optionally, the pipe segment type identification model is obtained by training based on a one-dimensional convolutional neural network model.

Optionally, the device for identifying bent and deformed pipe segments further includes a training module.

The training module is used to train the initial pipe segment type identification model based on the sample database and the labels corresponding to the sample data before inputting the second strain data corresponding to the pipe segment to be identified into the pipe segment type identification model to obtain the pipe segment type identification model; the sample database includes the pipe segment type , Absolute mileage range, girth weld number range, inspection date, pipe number and pipe segment strain value of the pipe segment strain data.

Optionally, the training module is specifically used for:

Obtain the sample data corresponding to the four types of anomalies in the pipe section, including sag, elbow, bending deformation and girth weld anomaly in the third strain data, and the sample database includes the sample data corresponding to the four types of pipe section anomalies.

The initial pipe segment type identification model is trained based on the sample data corresponding to the four pipe segment types and the labels corresponding to the sample data, and the pipe segment type identification model is obtained; among which, the labels corresponding to the sample data include 4 types of concave, elbow, bending deformation and abnormal girth weld Pipe segment exception type.

Optionally, the training module is specifically used for:

The sample data is input into the initial pipe segment type identification model, and after the feature processing of the input layer, convolution layer, pooling layer, flat layer, fully connected layer and output layer, the predicted type of the pipe segment is obtained.

According to the predicted type and label of the pipe segment, the network parameters of the initial pipe segment type identification model are updated to obtain the pipe segment type identification model.

Optionally, the training module is specifically used for:

According to the prediction type and label of the pipe segment, the loss function corresponding to the sample data is constructed.

According to the loss function corresponding to the sample data, the network parameters of the initial pipe segment type identification model are updated to obtain the pipe segment type identification model.

In a third aspect, the present application provides an electronic device, including: a memory and a processor;

Memory for storing computer programs.

The processor is configured to read the computer program stored in the memory, and execute the method for identifying the bent and deformed pipe segment of the first aspect according to the computer program in the memory.

In a fourth aspect, the present application provides a readable storage medium on which a computer program is stored, and a computer-executable instruction is stored in the computer program, and when the computer-executable instruction is executed by a processor, it is used to realize the bending deformation pipe segment as described above in the first aspect recognition methods.

In a fifth aspect, embodiments of the present application further provide a computer program product, including a computer program, which, when executed by a processor, implements the method for identifying a bent and deformed pipe segment of the first aspect.

In the method, device, electronic device and storage medium for identifying a bent and deformed pipe section provided by the present application, the first strain data obtained by the inertial measurement unit IMU measurement is obtained, and the first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline, According to the first strain data, a bending strain curve is obtained, a wave peak whose bending strain value is greater than a preset value in the bending strain curve is determined, and a pipe section to be identified is obtained. The pipe section to be identified is the pipe section corresponding to the preset interval on both sides of the wave crest in the pipeline, and the The second strain data corresponding to the pipe section to be identified is input into the pipe section type identification model, and the abnormal type of the pipe section of the pipe section to be identified is obtained. The method of the present application only needs to obtain the strain data of the IMU and determine the pipe segment to be identified, and input the second strain data corresponding to the pipe segment to be identified into the pipe segment type identification model, and then the abnormal type of the pipe segment of the pipe segment to be identified can be obtained without manual intervention. Judging the abnormal types of pipe sections can save time and manpower, and unify the judgment criteria, thereby improving the efficiency of judging abnormal types of pipe sections.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description serve to explain the principles of the application.

1 is a schematic flowchart of a method for identifying a bent and deformed pipe segment provided by an embodiment of the present application;

FIG. 2 is a schematic diagram of a pipe section to be identified according to an embodiment of the present application;

3 is a schematic structural diagram of a one-dimensional convolutional neural network provided by an embodiment of the present application;

4 is a schematic flowchart of another method for identifying a bent and deformed pipe segment provided by an embodiment of the present application;

5 is a schematic structural diagram of a device for identifying a bent and deformed pipe segment provided by an embodiment of the present application;

FIG. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Specific embodiments of the present application have been shown by the above-mentioned drawings, and will be described in more detail hereinafter. These drawings and written descriptions are not intended to limit the scope of the concepts of the present application in any way, but to illustrate the concepts of the present application to those skilled in the art by referring to specific embodiments.

Detailed ways

Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. Where the following description refers to the drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the illustrative examples below are not intended to represent all implementations consistent with this application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application as recited in the appended claims.

First, the terms involved in this application are explained:

Machine learning: A machine learning method is a method of constructing a machine learning model by training a large amount of data, and then using the model to classify and predict new data. Based on different learning forms, machine learning algorithms can be classified into supervised There are three types of learning, unsupervised learning and reinforcement learning. Machine learning involves multi-disciplinary interdisciplinary, mainly studies how computers simulate or realize the learning behavior of the human brain to acquire new knowledge or skills, reorganize the existing knowledge structure to continuously improve the performance of the model, so as to solve the optimal parameters .

Supervised learning: Use samples with known certain or certain characteristics as sample data, and train to obtain an optimal model to achieve data classification and prediction.

Feature engineering: In machine learning or statistics, also known as variable selection, attribute selection or variable subset selection, is the process of selecting relevant features and forming feature subsets in model building.

The technical solutions provided in the embodiments of the present application can be applied to the bending strain detection of buried oil and gas pipelines, in particular, to detect defects and anomalies on the pipelines by using the in-IMU detection technology. my country has formed the world's leading internal detection technology represented by IMU, magnetic flux leakage, etc. With the continuous popularization of pipeline internal detection technology, the obtained IMU strain data will also continue to grow, so the development of machine learning-based geological disasters is high risk. The research on intelligent identification of pipeline sections is of great significance for identifying large-strain pipeline sections under the influence of geological disasters and carrying out the integrity assessment of pipelines in geological disaster areas.

The existing method is based on the Matlab algorithm program to realize semi-automatic identification of high-risk pipeline sections of geological disasters, align the IMU strain data and geometric detection data of the same pipeline based on the girth weld number, and pre-find the bending in the IMU strain data through Matlab programming. All pipe sections with a strain greater than 0.125%, these pipe sections include elbows, depressions, bending deformation sections and abnormal girth weld pipes, mark the given mileage positions of elbows and depressions based on the aligned geometric inspection data, delete depressions, Strain data within the influence range of geometric feature points such as elbows, eliminating the interference of elbows and depressions. For the pipe sections with unmarked bending strain values greater than 0.125% in the remaining geometric detection, technicians with practical experience mark the pipe sections within a certain range on both sides of the large strain point as abnormal sections to be investigated by drawing pictures. According to the characteristics of the pipe sections within the range Determine the type of pipe segment by segment and label it, which are suspected elbow, suspected depression, abnormal girth weld segment and bending deformation segment. In this method, some pipe sections can be marked as elbows or depressions through the aligned geometric detection data, but there are still a large number of pipe sections that need to be judged manually, which requires a lot of time and energy, and requires the staff to have sufficient practice Experience, it is difficult for non-professional personnel to understand the difference in characteristics between different pipe types. In addition, due to the large base of data, there are still subtle feature differences between the same pipe segment types, and it is difficult to maintain a unified standard in the process of manual identification, so misjudgment is prone to occur. For ambiguous pipe segments characteristics, it is difficult to give an accurate judgment.

In order to solve the above problems, the present application proposes a bending deformation pipe section identification method. The first strain data is obtained by acquiring the first strain data measured by the inertial measurement unit IMU. The first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline. According to The first strain data, the bending strain curve is obtained, the wave crest whose bending strain value is greater than the preset value in the bending strain curve is determined, and the pipe section corresponding to the preset interval on both sides of the wave crest in the pipeline is the pipe section to be identified, and the first section corresponding to the pipe section to be identified is determined. The second strain data is input into the pipe segment type identification model, and the pipe segment abnormal type of the pipe segment to be identified is obtained. In this application, the pipe segment to be identified is determined according to the acquired first strain data measured by the IMU, and the second strain data corresponding to the pipe segment to be identified is input into the pipe segment type identification model, and the abnormal type of the pipe segment to be identified can be obtained. Manually judging the abnormal types of pipe sections can save time and manpower, and unify the judgment criteria, thereby improving the efficiency of judging abnormal types of pipe sections.

The technical solutions of the present application and how the technical solutions of the present application solve the above-mentioned technical problems will be described in detail below with specific embodiments. The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.

1 is a schematic flowchart of a method for identifying a bent and deformed pipe section according to an embodiment of the present application. The method for identifying a bent and deformed pipe section may be executed by software and/or a hardware device. For example, the hardware device may be an electronic device, such as a terminal or a server. By way of example, as shown in FIG. 1 , the method for identifying a bent and deformed pipe segment may include:

S101. Acquire first strain data measured by an inertial measurement unit IMU.

The first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline.

In this step, the inertial measurement unit (IMU) is a device that measures the three-axis attitude angle (or angular rate) and acceleration of the object. Its core components are three-dimensional orthogonal gyroscopes and accelerometers. The IMU is completing the measurement of the entire pipeline to be measured. After that, the data collected and recorded by the IMU is used for integration and other calculations, and the speed, position and attitude data information of the IMU at any time can be obtained. Based on the attitude data detected by the IMU, the deformation and strain state of the pipe section can be calculated. The first strain data can be understood as the IMU strain data obtained by measuring the pipeline to be measured by the in-IMU detection technology, including information such as the speed, position, and attitude data of the IMU at any time in the pipeline.

Specifically, the inertial measurement unit IMU disposed in the pipeline measures in the pipeline, and obtains the speed, position and attitude data of the IMU at any time in the pipeline, that is, the first strain data.

S102 , obtaining a bending strain curve according to the first strain data.

In this step, the bending strain curve can be expressed as a relation curve between the bending strain value and the absolute mileage, and the bending strain value can be calculated according to the IMU strain data.

Specifically, a corresponding bending strain value can be obtained according to the first strain data, and then a bending strain curve can be drawn according to the bending strain value and the corresponding absolute distance. Among them, the absolute mileage can be understood as the mileage from the strain point to the measurement starting position of the entire pipeline to be measured.

S103. Determine the peak of the bending strain curve with the bending strain value greater than the preset value, and obtain the pipe segment to be identified.

In this step, the pipe segment to be identified is the pipe segment corresponding to the preset interval on both sides of the wave crest in the pipeline.

The preset value may be set according to actual conditions or experience, for example, may be set to 0.125%, and the specific value of the preset value is not limited in this embodiment of the present application. When determining the peak of the bending strain value greater than the preset value in the bending strain curve, the peak can be determined by using the findpeak function in python, and it can also be realized by programming. The findpeak function can be expressed as:

peaks=scipy.signal.find_peaks(x,threshold,distance) (1)

In formula (1): x is a signal with a peak value, threshold is the set recognition threshold, and distance is the minimum horizontal distance between two peaks.

Specifically, according to the set preset value, a wave peak whose bending strain value is greater than the preset value in the bending strain curve can be found, and then two strains whose bending strain value is the preset value on the bending strain curve on the left and right sides of the wave peak are respectively found. The absolute mileage interval corresponding to these two strain points is the preset interval on both sides of the wave crest, and then the pipe segment within the absolute mileage interval in the pipeline can be found, that is, the pipe segment to be identified.

Among them, Fig. 2 is a schematic diagram of a pipe segment to be identified. As shown in Fig. 2, the preset value is 0.125%. According to the preset value of 0.125%, the peaks A and B with the bending strain value greater than 0.125% in the bending strain curve can be found. and C, taking the wave peak A as an example, the two strain points with a bending strain value of 0.125% on the bending strain curve on the left and right sides of the wave peak A are E and F respectively, the absolute mileage corresponding to the strain point E is 3330m, and the strain point F corresponds to The absolute mileage is 3340m, then the absolute mileage interval corresponding to the two strain points is (3330, 3340), and then the preset interval (3330, 3340) on both sides of the peak A can be obtained. At this time, the preset interval (3330, 3340 ) corresponds to the pipeline segment 1 to be identified.

S104: Input the second strain data corresponding to the pipe section to be identified into the pipe section type identification model to obtain the abnormal type of the pipe section of the pipe section to be identified.

In this step, the second strain data can be understood as the IMU strain data obtained by measuring the pipe segment to be identified through the in-IMU detection technology.

Exemplarily, the pipe segment type identification model may be obtained by training based on a one-dimensional neural network model, such as a one-dimensional convolutional neural network model. The initial pipe segment type identification model is trained through a one-dimensional neural network model, and feature engineering is not required to extract feature values, which can reduce the amount of computation and simplify the feature extraction process.

The pipe segment type identification model can also be obtained by training other network models, such as a multi-dimensional convolutional neural network model, a recurrent neural network model, or a random forest model.

Exemplarily, before the second strain data corresponding to the pipe segment to be identified is input into the pipe segment type identification model, the initial pipe segment type identification model is trained first, and the specific steps may be: training the initial pipe segment type identification model based on the label corresponding to the sample database and the sample data, Obtain the model for identifying the type of pipe segment.

Among them, the sample database includes the pipe segment type, the absolute mileage range where the pipe segment strain data is located, the girth weld number range, the detection date, the pipe number and the pipe segment strain value. The sample database can be Mysql database.

Specifically, when training the initial pipe segment type identification model, the sample data in the sample database and the labels corresponding to the sample data are input into the initial pipe segment type identification model, and the pipe segment type is obtained by training according to the sample data in the sample database and the labels corresponding to the sample data. Identify the model.

Through years of pipeline data monitoring practice, a large amount of pipeline IMU strain data has been obtained. Although there is a large amount of IMU strain data, there is still a lack of corresponding data processing methods. In this scheme, a machine learning method driven by big data is formed by establishing a sample database of different types of pipe sections, and with the popularization and application of in-pipe detection technology, the database is continuously expanded, which can counteract the machine learning method and improve the types of pipe sections. recognition accuracy.

Exemplarily, the sample database includes sample data corresponding to four types of pipe section anomalies: sag, elbow, bending deformation and girth weld anomaly. When the initial pipe segment type identification model is trained according to the sample data in the sample database and the labels corresponding to the sample data, the sample data corresponding to the four types of pipe segment anomalies in the third strain data: sag, elbow, bending deformation and girth weld anomaly are obtained. Based on the sample data corresponding to the four pipe types and the labels corresponding to the sample data, the initial pipe type identification model is trained, and the pipe type identification model is obtained.

Among them, the labels corresponding to the sample data include 4 types of pipe section anomalies: depression, elbow, bending deformation and girth weld anomaly. The third strain data can be understood as historical IMU strain data obtained based on the detection technology in the IMU, which can be obtained through a local database or downloaded from a server.

Exemplarily, Table 1 is a sample database. As shown in Table 1, the sample database includes the pipe segment type, the absolute mileage range where the pipe segment strain data is located, the girth weld number range, the detection date, the pipe number and the pipe segment strain value. Among them, the pipe segment types include 4 types of pipe segment anomalies, including sag, elbow, bending deformation and girth weld anomaly, and the four types of sag, elbow, bending deformation and girth weld anomaly are labeled with numbers 1, 2, and 1, respectively. 3 and 4 indicate. Take the pipe segment type numbered 1 as an example, the pipe segment type is sag, the corresponding label is 1, the absolute mileage range is 600.9-620.5m, the girth weld number range is 400-420, the inspection date is May 2016, the pipeline The number is segment A, and the strain value of the pipe segment is [-0.86%, . . . , 0.23%].

Table 1 Sample database

Specifically, firstly, the sample data corresponding to four types of pipe section anomalies, including sag, elbow, bending deformation, and girth weld anomaly, are obtained from the historical IMU strain data. Among them, sample data corresponding to depressions and elbows in the third strain data are extracted by a geometric detection method, and sample data corresponding to bending deformation and girth weld abnormalities in the third strain data are extracted by a manual identification method. According to the sample data corresponding to the four types of pipe segments and the labels corresponding to the sample data, the initial pipe segment type identification model is trained, and the pipe segment type identification model is obtained.

In this scheme, through the establishment of a sample database of different types of pipe sections, a data foundation is laid for the application and continuous improvement of machine learning methods. Based on the existing database, the detailed information of the required pipe sections can be quickly found, and it can also be used for the follow-up of the pipe sections. Development provides a basis for judgment. In addition, the identification model of the pipe segment type obtained by training the sample data and labels corresponding to the four types of abnormal types of pipe segment, including sag, elbow, bending deformation and girth weld abnormality, only needs to provide geometric detection data and manual labor when establishing the sample database. As a result of the identification, after training the identification model for the type of pipe segment, the type of pipe segment can be predicted based only on the IMU strain data, and no manual identification method is needed to judge the abnormal type of the pipe segment. Type judgment efficiency.

Exemplarily, the pipe segment type identification model may be obtained by training based on a one-dimensional convolutional neural network model, and the model may include six parts: an input layer, a convolutional layer, a pooling layer, a flat layer, a fully connected layer, and an output layer, When training the initial pipe segment type identification model, the sample data is first input into the initial pipe segment type identification model, and after the feature processing of the input layer, convolution layer, pooling layer, flat layer, fully connected layer and output layer, we get Pipe segment prediction type, update the network parameters of the initial pipe segment type identification model according to the pipe segment prediction type and label, and obtain the pipe segment type identification model.

Specifically, the tensorflow module in python can be used to build the basic framework of the one-dimensional convolutional neural network model. Figure 3 is a schematic diagram of the one-dimensional convolutional neural network structure. The model can include an input layer (not shown), a convolution layer, Pooling layer, flattening layer, fully connected layer and output layer (not shown) 6 parts.

The input layer of a one-dimensional convolutional neural network model can receive one-dimensional or two-dimensional arrays. Input sample data in the input layer of the initial pipe segment type identification model. The sample data is the IMU strain data, including the components in the horizontal and vertical directions, that is, the 2×n pipe segment strain characteristic matrix, which can be expressed as:

为第n个应变点的水平应变分量；

为第n个应变点的垂直应变分量。In formula (2): T is the strain value matrix;

is the horizontal strain component of the nth strain point;

is the vertical strain component of the nth strain point.

The input layer is the sample data input port of the neural network model. The preset input dimension of this layer needs to be consistent with the dimension of the sample data. The one-dimensional convolutional neural network needs to convert the multi-dimensional sample data, that is, the second strain data corresponding to 2 in the input layer. The ×n pipe segment strain characteristic matrix is converted into one-dimensional data.

The convolutional layer is a feature extraction layer, which reduces the dimension of the original sample data through the convolution operation, and extracts the main features of the sample at the same time to prevent the model from overfitting due to too many parameters.

The function of the pooling layer is to further reduce the size of the feature data, which can greatly reduce the number of parameters in the network and remove redundant information, mainly including mean sampling and maximum sampling.

Due to the filter, the data dimension is changed, and the function of the flat layer is to convert the data to be processed into a one-dimensional vector before inputting to the neural network of the fully connected layer, which corresponds to the neural unit of the fully connected layer.

The fully connected layer is composed of multiple neurons, and each neuron is fully connected, and the local features sent from the pooling layer are reassembled through the weight matrix to form a complete global feature.

The output layer is composed of a regression classifier, which can use the softmax function to map the outputs of multiple neurons into a fixed interval, so as to realize the identification of various types of abnormal types of pipe segments of the pipe segment type identification model.

After the feature processing of the input layer, convolution layer, pooling layer, flat layer, fully connected layer and output layer, the prediction type of the pipe segment is obtained. The spool type identifies the model.

In this solution, the initial pipe segment type identification model is trained by the one-dimensional convolutional neural network model, and feature engineering is not required to extract feature values, which can reduce the amount of computation and simplify the feature extraction process.

Exemplarily, when updating the network parameters of the initial pipe segment type identification model according to the pipe segment prediction type and label, first, construct a loss function corresponding to the sample data according to the pipe segment prediction type and label, and update the initial pipe segment according to the loss function corresponding to the sample data. The network parameters of the type identification model are obtained to obtain the pipe segment type identification model.

Specifically, when the loss function corresponding to the sample data is constructed according to the prediction type and label of the pipe segment, the loss function can be a cross entropy loss function, and the formula can be expressed as:

In formula (3): n is the number of samples, x is the sample number, y is the real value, and a is the predicted value.

Exemplarily, divide all the sample data in the sample database into a training data set and a test data set according to a certain ratio, for example, set according to 4:1, and set the batch size batch and the number of iterations epoch in the training process, Store the trained pipe segment type recognition model and its parameters.

In this scheme, by constructing the loss function corresponding to the predicted type of the pipe segment and the label corresponding to the sample data, the network parameters of the initial segment type identification model are continuously updated, so that the accuracy of the training of the pipe segment type identification model can be improved.

When evaluating the comprehensive performance of the model, the parameters of the network model are tuned to determine the optimal hyperparameters. In the classification supervised learning model, the commonly used classification model evaluation indicators include accuracy rate, precision rate, recall rate and FI value, etc. The formula can be expressed as:

In the above formula, Accuracy represents the accuracy rate, Precision represents the precision rate, Recall represents the recall rate, FIScore represents the FI value, TP represents the correct classification as a positive sample, TN represents the correct classification as a negative sample, FP represents an incorrect classification as a positive sample, FN represents Misclassified as negative samples.

Exemplarily, when the performance evaluation of the pipe segment type identification model is performed, when the accuracy rate of the pipe segment type identification model is greater than or equal to a preset value, the pipe segment type identification model can accurately identify the abnormal type of the pipe segment.

In the method for identifying a bent and deformed pipe segment provided by the embodiment of the present application, the first strain data obtained by the inertial measurement unit (IMU) measurement is obtained, and the first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline. According to the first strain data, Obtain the bending strain curve, determine the peak of the bending strain curve with the bending strain value greater than the preset value, and the preset interval on both sides of the wave crest in the pipeline corresponding to the pipeline segment to be identified, and input the second strain data corresponding to the pipeline segment to be identified into the pipeline segment The type identification model is used to obtain the abnormal type of the pipe segment to be identified. In the present application, the pipe section to be identified is determined according to the acquired IMU strain data, and the second strain data corresponding to the pipe section to be identified is input into the pipe section type identification model, and the abnormal type of the pipe section to be identified can be obtained, and no manual operation of the abnormal type of the pipe section is required. The judgment can save time and manpower, and unify the judgment standard, thereby improving the judgment efficiency of the abnormal type of the pipe section.

FIG. 4 is a schematic flowchart of another method for identifying a bent and deformed pipe section provided by an embodiment of the present application. Referring to FIG. 4 , the method for identifying a bent and deformed pipe section may include a training process and an application process:

When training the initial pipe segment type identification model, the third strain data is first preprocessed, that is, the sample data corresponding to the concave and elbow in the third strain data are extracted by the geometric detection method, and the third strain data is extracted by the manual identification method. Sample data corresponding to bending deformation and girth weld abnormality in the strain data; then build a sample database based on the sample data corresponding to the four types of pipe segments extracted, and train the initial pipe segment type identification model based on the labels corresponding to the sample database and the sample data, and obtain the pipe segment Type identification model; finally, perform performance evaluation on the pipe segment type identification model. When the accuracy rate of the pipe segment type identification model is greater than or equal to the preset value, the pipe segment type identification model can accurately identify the abnormal type of the pipe segment.

When using the pipe segment type identification model to identify the abnormal type of the pipe segment, first obtain the first strain data measured by the inertial measurement unit IMU. The first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline. According to the first strain data , obtain the bending strain curve, determine the peak of the bending strain value in the bending strain curve greater than the preset value, obtain the pipe section to be identified, input the second strain data corresponding to the pipe section to be identified into the pipe section type identification model, and obtain the abnormal type of the pipe section of the pipe section to be identified , so as to realize the intelligent identification and classification of high-risk sections of geological disasters.

The bending deformation pipe section identification method provided in the embodiment of the present application is a pipe section type identification model obtained by training the sample data and labels corresponding to the four types of pipe section abnormality: sag, elbow, bending deformation and girth weld abnormality, and obtains the inertial measurement unit IMU. The first strain data obtained by measurement, the first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline, the bending strain curve is obtained according to the first strain data, and it is determined that the bending strain value in the bending strain curve is greater than the preset value. The wave crest, the pipe section corresponding to the preset interval on both sides of the wave crest in the pipeline is the pipe section to be identified, and the second strain data corresponding to the to-be-identified pipe section is input into the pipe section type identification model to obtain the pipe section abnormality type of the to-be-identified pipe section. In this application, it is necessary to provide geometric detection data and manual identification results when establishing a sample database. After training to obtain a pipe segment type identification model, the type of pipe segment abnormality can be predicted based only on the IMU strain data, and no manual identification method is required to detect the pipe segment abnormality. The type judgment can save time and manpower, and unify the judgment standards, thereby improving the judgment efficiency of abnormal types of pipe sections.

FIG. 5 is a schematic structural diagram of a bending and deformed pipe segment identification device 50 provided by an embodiment of the present application. For example, please refer to FIG. 5 . The bending and deformed pipe segment identification device 50 includes:

The acquiring module 501 is configured to acquire first strain data measured by the inertial measurement unit IMU, where the first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline.

The obtaining module 501 is further configured to obtain a bending strain curve according to the first strain data.

The determination module 502 is used to determine the peaks in the bending strain curve with the bending strain value greater than the preset value, and obtain the pipe sections to be identified, which are the pipe sections corresponding to the preset intervals on both sides of the wave crests in the pipeline.

The identification module 503 is configured to input the second strain data corresponding to the pipe section to be identified into the pipe section type identification model to obtain the abnormal type of the pipe section of the pipe section to be identified.

Optionally, the pipe segment type identification model is obtained by training based on a one-dimensional convolutional neural network model.

Optionally, the apparatus for identifying bent and deformed pipe segments further includes a training module 504 .

The training module 504 is used to train the initial pipe segment type identification model based on the sample database and the labels corresponding to the sample data before inputting the second strain data corresponding to the pipe segment to be identified into the pipe segment type identification model to obtain the pipe segment type identification model; the sample database includes the pipe segment type identification model. Type, absolute mileage range where the pipe segment strain data is located, girth weld number range, inspection date, pipe number and pipe segment strain value.

Optionally, the training module 504 is specifically used for:

Obtain the sample data corresponding to the four types of anomalies in the pipe section, including sag, elbow, bending deformation and girth weld anomaly in the third strain data, and the sample database includes the sample data corresponding to the four types of pipe section anomalies.

The initial pipe segment type identification model is trained based on the sample data corresponding to the four pipe segment types and the labels corresponding to the sample data, and the pipe segment type identification model is obtained; among which, the labels corresponding to the sample data include 4 types of concave, elbow, bending deformation and abnormal girth weld Pipe segment exception type.

Optionally, the training module 504 is specifically used for:

The sample data is input into the initial pipe segment type identification model, and after the feature processing of the input layer, convolution layer, pooling layer, flat layer, fully connected layer and output layer, the predicted type of the pipe segment is obtained.

According to the predicted type and label of the pipe segment, the network parameters of the initial pipe segment type identification model are updated to obtain the pipe segment type identification model.

Optionally, the training module 504 is specifically used for:

According to the prediction type and label of the pipe segment, the loss function corresponding to the sample data is constructed.

According to the loss function corresponding to the sample data, the network parameters of the initial pipe segment type identification model are updated to obtain the pipe segment type identification model.

The bending deformation pipe segment identification device 50 shown in the embodiment of the present application can implement the technical solution of the bending deformation pipe segment identification method in the above-mentioned embodiment, and its implementation principle and beneficial effects are similar to the implementation principle and beneficial effects of the bending deformation pipe segment identification method. Reference can be made to the realization principle and beneficial effects of the method for identifying the bent and deformed pipe segment, which will not be repeated here.

FIG. 6 is a schematic structural diagram of an electronic device 60 provided by an embodiment of the present application. For an example, please refer to FIG. 6 , the electronic device 60 may include a processor 601 and a memory 602; wherein,

The memory 602 is used to store computer programs.

The processor 601 is configured to read the computer program stored in the memory 602, and execute the method for identifying the bent and deformed pipe segment in the above embodiment according to the computer program in the memory 602.

Optionally, the memory 602 may be independent or integrated with the processor 601 . When the memory 602 is a device independent of the processor 601 , the electronic device 60 may further include: a bus for connecting the memory 602 and the processor 601 .

Optionally, this embodiment further includes: a communication interface, where the communication interface can be connected to the processor 601 through a bus. The processor 601 can control the communication interface to realize the above-mentioned functions of obtaining and sending of the electronic device 60 .

For example, in this embodiment of the present application, the electronic device 60 may be a terminal or a server, which may be specifically set according to actual needs.

The electronic device 60 shown in the embodiment of the present application can implement the technical solution of the method for identifying a bent and deformed pipe segment in the above-mentioned embodiments, and its implementation principle and beneficial effects are similar to those of the method for identifying a bent and deformed pipe segment. The realization principle and beneficial effects of the deformed pipe segment identification method will not be repeated here.

Embodiments of the present application further provide a computer-readable storage medium, where computer-executable instructions are stored in the computer-readable storage medium, and when a processor executes the computer-executable instructions, the method for identifying a bent and deformed pipe segment in the foregoing embodiment is implemented. The implementation principle and beneficial effects of the technical solution are similar to those of the method for identifying bent and deformed pipe sections. Please refer to the implementation principle and beneficial effects of the method for identifying curved and deformed pipe sections, which will not be repeated here.

Embodiments of the present application further provide a computer program product, including a computer program, when the computer program is executed by a processor, the technical solution for realizing the method for identifying a bending deformation pipe segment in the above embodiment, its realization principle, beneficial effects and bending deformation The realization principle and beneficial effects of the pipe segment identification method are similar, and reference may be made to the realization principle and beneficial effects of the bending deformation pipe segment identification method, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and components shown as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment. In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware, or may be implemented in the form of hardware plus software functional units.

The above-mentioned integrated modules implemented in the form of software functional modules may be stored in a computer-readable storage medium. The above-mentioned software function modules are stored in a storage medium, and include several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (processor) to perform some steps of the methods of the various embodiments of the present application .

It should be understood that the above-mentioned processor may be a central processing unit (Central Processing Unit, CPU), and may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC) etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in conjunction with the invention can be directly embodied as executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

The memory may include high-speed random access memory (Random Access Memory, RAM), and may also include non-volatile memory (Non-Volatile Memory, NVM), such as at least one disk Read memory, magnetic disk or CD, etc.

The bus may be an industry standard architecture (Industry Standard Architecture, ISA) bus, a Peripheral Component (Peripheral Component, PCI) bus, or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, or the like. The bus can be divided into address bus, data bus, control bus and so on. For convenience of representation, the buses in the drawings of the present application are not limited to only one bus or one type of bus.

The above-mentioned computer-readable storage medium can be realized by any type of volatile or non-volatile storage device or their combination, such as static random-access memory (Static Random-Access Memory, SRAM), electrically erasable programmable Read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), only Read-Only Memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. A storage medium can be any available medium that can be accessed by a general purpose or special purpose computer.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: The technical solutions described in the foregoing embodiments can still be modified, or some or all of the technical features thereof can be equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present application. scope.

Claims (10)

1. a bending deformation pipe section identification method, is characterized in that, comprises: acquiring first strain data measured by the inertial measurement unit IMU, where the first strain data is pipeline strain data obtained when the IMU completes the detection in the pipeline; obtaining a bending strain curve according to the first strain data; determining a wave crest with a bending strain value greater than a preset value in the bending strain curve, and obtaining a pipe section to be identified, where the pipe section to be identified is a pipe section corresponding to a preset interval on both sides of the wave crest in the pipeline; Inputting the second strain data corresponding to the pipe section to be identified into the pipe section type identification model to obtain the pipe section abnormality type of the pipe section to be identified. 2 . The method according to claim 1 , wherein the pipe segment type identification model is obtained by training based on a one-dimensional convolutional neural network model. 3 . 3. The method according to claim 1 or 2, wherein before the inputting the second strain data corresponding to the pipe segment to be identified into the pipe segment type identification model, the method comprises: The initial pipe segment type identification model is trained based on the sample database and the labels corresponding to the sample data, and the pipe segment type identification model is obtained; the sample database includes the pipe segment type, the absolute mileage range where the pipe segment strain data is located, the girth weld number range, and the detection date. , pipe number and segment strain value. 4. The method according to claim 3, wherein the initial pipe segment type identification model is trained based on the label corresponding to the sample database and the sample data to obtain the pipe segment type identification model, comprising: Obtain sample data corresponding to the four types of pipe section anomalies in the third strain data: sag, elbow, bending deformation and girth weld anomaly, and the sample database includes sample data corresponding to the four types of pipe section anomalies; The initial pipe segment type identification model is trained based on the sample data corresponding to the four pipe segment types and the labels corresponding to the sample data, and the pipe segment type identification model is obtained; wherein the labels corresponding to the sample data include depressions, elbows 4 types of pipe section anomalies, bending deformation and girth weld anomaly. 5 . The method according to claim 4 , wherein the initial pipe segment type identification model is trained based on the sample data corresponding to the four pipe segment types and the labels corresponding to the sample data to obtain the pipe segment type. 6 . Identify models, including: Inputting the sample data into the initial pipe segment type identification model, and after feature processing of the input layer, the convolution layer, the pooling layer, the flat layer, the fully connected layer and the output layer, the pipe segment prediction type is obtained; According to the predicted type of the pipe segment and the label, the network parameters of the initial pipe segment type identification model are updated to obtain the pipe segment type identification model. 6. The method according to claim 5, wherein, according to the predicted type of the pipe segment and the label, updating the network parameters of the initial pipe segment type identification model to obtain the pipe segment type identification model, comprising: Constructing a loss function corresponding to the sample data according to the pipe segment prediction type and the label; According to the loss function corresponding to the sample data, the network parameters of the initial pipe segment type identification model are updated to obtain the pipe segment type identification model. 7. A bending deformation pipe segment identification device, characterized in that, comprising: an acquisition module, configured to acquire first strain data measured by the inertial measurement unit IMU, where the first strain data is the pipeline strain data obtained when the IMU completes the detection in the pipeline; The obtaining module is further configured to obtain a bending strain curve according to the first strain data; A determination module, configured to determine a wave crest whose bending strain value is greater than a preset value in the bending strain curve, and obtain a pipe segment to be identified, where the pipe segment to be identified is the pipe segment corresponding to the preset interval on both sides of the wave crest in the pipeline ; The identification module is configured to input the second strain data corresponding to the pipe section to be identified into the pipe section type identification model to obtain the abnormal type of the pipe section of the pipe section to be identified. 8. An electronic device, comprising: a memory and a processor; the memory for storing computer programs; The processor is configured to read the computer program stored in the memory, and execute the method for identifying a bent and deformed pipe segment according to any one of claims 1-6 according to the computer program in the memory. 9. A readable storage medium on which a computer program is stored, characterized in that the computer program stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, is used to achieve the above-mentioned claims 1- 6. The method for identifying a bent and deformed pipe segment according to any one of them. 10. A computer program product, comprising a computer program, which, when executed by a processor, is used to implement the method for identifying a bent and deformed pipe segment according to any one of the preceding claims 1-6.

Images