Rapid detection method for tensile property of PE gas pipeline

Abstract

The invention relates to the technical field of pipeline detection, and discloses a method for rapidly detecting the tensile property of a PE gas pipeline, which comprises the following steps: installing an ultrasonic transducer on the surface of the PE gas pipeline, exciting ultrasonic guided waves, collecting signals, and extracting the characteristics of elasticity modulus, wave velocity and the like of the pipeline through time-frequency analysis; and then analyzing the signal through a deep learning model (CNN + LSTM), and constructing a tensile strength prediction model. Compared with a traditional destructive test, the method can rapidly and nondestructively predict the tensile strength of the pipeline on the construction site, has high precision (+ / -5%), and realizes accurate evaluation of the state of the pipeline in a complex environment. The deep learning model can process pipeline data under various specifications, materials and climate conditions, wide adaptability and high stability of the method are guaranteed, and the working efficiency and the safety of pipeline detection are greatly improved. Meanwhile, the prediction value is dynamically optimized through an error correction mechanism, the result precision is ensured, and the prediction reliability is enhanced.

Description

Technical Field

[0001] The invention relates to the technical field of pipeline performance detection, and in particular to a rapid detection method for the tensile performance of a PE gas pipeline. Background Art

[0002] PE gas pipeline is an important part of the city gas transmission system, and its long-term stable operation is directly related to public safety. The tensile strength of the pipeline is an important parameter for judging its structural safety. Usually, its service life is evaluated by testing the tensile properties of the pipeline material. However, traditional testing methods are mostly destructive tests. Although these methods are accurate, they have the following problems: 1. Destructive testing: Traditional tensile strength testing requires sampling of the pipeline and measuring the tensile strength through physical tensile testing. This method not only causes pipeline damage, but also fails to monitor the pipeline in real time after installation, resulting in high costs and time delays for subsequent maintenance and repairs. At the same time, sampling and destructive testing cannot conduct large-scale rapid inspections of in-service pipelines, limiting their popularity in practical applications.

[0003] 2. Limitations of ultrasonic guided wave technology: In order to solve the problems of traditional destructive testing, the industry has gradually adopted ultrasonic guided wave technology for non-destructive testing. Ultrasonic guided wave technology can effectively detect defects or performance changes inside the pipeline by stimulating ultrasonic signals in the pipeline and monitoring the propagation characteristics of the signal. However, most existing ultrasonic guided wave detection methods rely on manual analysis, and there is a large degree of subjectivity and uncertainty in the signal processing process, which limits the accuracy of the results. At the same time, this method has poor adaptability in complex environments, especially under the influence of different pipeline specifications, material types and environmental factors, its effect is difficult to guarantee.

[0004] 3. Challenges in the application of artificial intelligence technology: In recent years, artificial intelligence has been applied to the processing and analysis of ultrasonic guided wave signals. Through technologies such as deep learning, the degree of automation and accuracy of signal processing can be improved. However, in the training process of deep learning models in existing artificial intelligence-based detection technologies, the training data usually used is mostly standard data in laboratory environments, and it fails to effectively combine the actual use data of on-site pipelines, resulting in poor generalization and adaptability of the model. In addition, the models in existing technologies are not adaptable enough to various environmental factors (such as temperature, humidity, pipeline aging, etc.), making it difficult to provide high-precision and stable prediction results in practical applications. Summary of the invention

[0005] The technical problem solved by the present invention is to provide a rapid detection method for the tensile performance of a PE gas pipeline in view of the defects existing in the above-mentioned prior art, so as to solve the problems raised in the above-mentioned background technology.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows: a rapid detection method for the tensile performance of a PE gas pipeline, characterized in that the method comprises the following steps: Step 1, installing at least one ultrasonic transducer on the surface of the PE gas pipeline, for exciting ultrasonic guided waves to propagate along the axial direction of the pipeline and excite stress waves inside the pipeline, wherein the ultrasonic transducer is a piezoelectric transducer or a laser transducer, and the excitation frequency of the ultrasonic wave is between 10kHz and 500kHz, and the ultrasonic signal can propagate along the axial direction of the pipeline and interact with the stress waves inside the pipeline; Step 2, using at least one ultrasonic receiving sensor to collect ultrasonic reflection signals and transmission signals after propagation through the pipeline, and recording the propagation characteristics of the ultrasonic signal in real time, wherein the propagation characteristics of the ultrasonic signal include signal attenuation, frequency shift, propagation time and reflection mode parameters; Step 3, extracting features of the collected ultrasonic signal through the signal processing module, performing time-frequency analysis on the signal using wavelet transform or Fourier transform, and extracting the elastic modulus, wave velocity, attenuation factor and frequency characteristics of the pipeline as candidate characteristic parameters of the pipeline tensile strength; Step 4: Based on the extracted candidate feature parameters, a deep learning algorithm is used to analyze the ultrasonic signal through a convolutional neural network (CNN) and a long short-term memory network (LSTM) to construct a pipeline tensile strength prediction model, wherein the prediction model is obtained by training with historical experimental data, and the historical experimental data includes standard tensile strength data of PE pipelines under different materials, thicknesses, strain rates and environmental conditions; Step 5, using the prediction model to predict the ultrasonic signal in real time and calculate the tensile strength of the pipeline; Step 6: If the tensile strength prediction result is lower than the set threshold, a maintenance suggestion is automatically generated and transmitted to the monitoring platform or on-site operators through the communication interface for subsequent maintenance operations to avoid possible risks of pipeline rupture or gas leakage.

[0007] As a further solution of the present invention, in step 5, the ultrasonic signal is predicted and calculated in real time by the prediction model to calculate the tensile strength of the pipeline, which specifically includes the following sub-steps: s1, input data: the ultrasonic signal characteristic parameters extracted in step 3 are used as input data and passed into the tensile strength prediction model; wherein the ultrasonic signal characteristic parameters include elastic modulus , wave speed , attenuation factor and frequency characteristics ; s2, feature extraction and processing: The convolutional neural network (CNN) is used to deeply extract time domain and frequency domain features, and the long short-term memory network (LSTM) is used to model the time series data of the signal, so as to further learn the elastic behavior of the pipeline and the dynamic changes of its tensile strength; and the weights of the neural network are updated by optimizing the back-propagation algorithm in this process; s3, tensile strength prediction model calculation: Use the trained CNN and LSTM models to analyze and calculate the input signal features to obtain the tensile strength prediction value of the pipeline. Calculated by the following formula:

[0008] in: It is the weight coefficient matrix obtained through historical data training, which is used for weight optimization and correction of prediction results; s4, weight coefficient adjustment: by comparing the error between historical experimental data and predicted values, the least square method or other appropriate regression algorithms are used to dynamically adjust the weight coefficients in the prediction model. To ensure the accuracy of the predicted results and their consistency with the actual tensile strength; s5, tensile strength output: output the tensile strength prediction value of the adjusted model It serves as an indicator of the current tensile performance of the pipeline and makes subsequent maintenance decisions.

[0009] As a further aspect of the present invention, the frequency range of the ultrasonic signal is 20 kHz to 300 kHz to increase the sensitivity to the pipe material and enhance the signal penetration.

[0010] As a further solution of the present invention, the ultrasonic transducer is a piezoelectric transducer, and the piezoelectric transducer is made of PZT lead titanate material, has high sensitivity and is adaptable to working requirements in high temperature environments.

[0011] As a further solution of the present invention, the signal processing module includes a time domain analysis unit and a frequency domain analysis unit. The time domain analysis unit is used to extract the time characteristics of the signal, and the frequency domain analysis unit is used to extract the frequency characteristics of the signal to further optimize the feature extraction of the signal.

[0012] As a further solution of the present invention, the deep learning model is a convolutional neural network CNN, and the CNN model includes at least three convolutional layers, a pooling layer and a fully connected layer to achieve deep learning of signal features and accurate prediction of tensile properties.

[0013] As a further solution of the present invention, the training data of the prediction model is derived from multivariate experimental data including different ambient temperatures, humidity, pipeline materials and manufacturing process parameters, and data consistency is ensured through data normalization processing.

[0014] As a further solution of the present invention, the deep learning algorithm is trained and optimized by the gradient descent method, and the cross-validation method is used to verify the generalization ability of the model to reduce overfitting.

[0015] As a further solution of the present invention, the tensile strength prediction result is combined with the actual use environment conditions of the pipeline, and risk assessment is performed by setting different thresholds. If the prediction result is lower than the safety threshold, maintenance suggestions are automatically generated and transmitted to relevant operators through the communication interface.

[0016] As a further solution of the present invention, the collection of ultrasonic reflection signals and transmission signals includes preprocessing the signals to remove background noise, and the noise removal method includes but is not limited to bandpass filter and wavelet transform to ensure high signal quality.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: by combining ultrasonic guided wave technology with deep learning algorithms, rapid non-destructive testing of the tensile properties of PE gas pipelines is achieved. Not only does it not require destructive sampling of the pipeline, but the tensile strength of the pipeline can also be directly predicted through the ultrasonic signals collected in real time. The deep learning model (CNN+LSTM) is used to accurately analyze the time-frequency characteristics of the signal, so that the prediction accuracy can reach ±5%, and it has strong adaptability to complex environments, and can quickly complete the evaluation of tensile properties at the construction site, greatly improving work efficiency and the safety of pipeline detection; the ultrasonic signal is predicted in real time through the deep learning model, and the error correction mechanism is used to correct the predicted value of tensile strength in combination with standard experimental data, ensuring the high accuracy of the prediction result. The error correction method dynamically adjusts the predicted value through optimization algorithms such as the least squares method to ensure that the error between the predicted value of tensile strength and the true value is always controlled within a predetermined range; and it can adapt to a variety of specifications and environmental conditions, is suitable for PE gas pipelines of different thicknesses and different material grades, and can effectively handle the prediction of pipeline tensile strength under various climates and strain rates. Since the deep learning model takes into account a large amount of historical experimental data and a variety of experimental environments during the training process, this method has a high adaptability to various environmental changes and different types of pipelines, and can maintain high accuracy and high stability in a wider range of practical application scenarios; as well as being able to adapt to a variety of specifications and environmental conditions, it is suitable for PE gas pipelines of different thicknesses and different material grades, and can effectively handle the prediction of pipeline tensile strength under various climates and strain rates. Since the deep learning model takes into account a large amount of historical experimental data and a variety of experimental environments during the training process, this method has a high adaptability to various environmental changes and different types of pipelines, and can maintain high accuracy and high stability in a wider range of practical application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 creative labor.

[0019] Figure 1 It is a schematic diagram of the structure of the PE gas pipeline tensile performance detection system of the present invention. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] See also Figure 1 In an embodiment of the present invention, a rapid detection method for the tensile performance of a PE gas pipeline is provided, characterized in that the method comprises the following steps: Step 1, installing at least one ultrasonic transducer on the surface of the PE gas pipeline, for exciting ultrasonic guided waves to propagate along the axial direction of the pipeline and excite stress waves inside the pipeline, wherein the ultrasonic transducer is a piezoelectric transducer or a laser transducer, and the excitation frequency of the ultrasonic wave is between 10kHz and 500kHz, and the ultrasonic signal can propagate along the axial direction of the pipeline and interact with the stress waves inside the pipeline; Step 2, using at least one ultrasonic receiving sensor to collect ultrasonic reflection signals and transmission signals after propagation through the pipeline, and recording the propagation characteristics of the ultrasonic signal in real time, wherein the propagation characteristics of the ultrasonic signal include signal attenuation, frequency shift, propagation time and reflection mode parameters; Step 3, extracting features of the collected ultrasonic signal through the signal processing module, performing time-frequency analysis on the signal using wavelet transform or Fourier transform, and extracting the elastic modulus, wave velocity, attenuation factor and frequency characteristics of the pipeline as candidate characteristic parameters of the pipeline tensile strength; Step 4: Based on the extracted candidate feature parameters, a deep learning algorithm is used to analyze the ultrasonic signal through a convolutional neural network (CNN) and a long short-term memory network (LSTM) to construct a pipeline tensile strength prediction model, wherein the prediction model is obtained by training with historical experimental data, and the historical experimental data includes standard tensile strength data of PE pipelines under different materials, thicknesses, strain rates and environmental conditions; Step 5, using the prediction model to predict the ultrasonic signal in real time and calculate the tensile strength of the pipeline; Step 6: If the tensile strength prediction result is lower than the set threshold, a maintenance suggestion is automatically generated and transmitted to the monitoring platform or on-site operators through the communication interface for subsequent maintenance operations to avoid possible risks of pipeline rupture or gas leakage.

[0022] As a further solution of the present invention, in step 5, the ultrasonic signal is predicted and calculated in real time by the prediction model to calculate the tensile strength of the pipeline, which specifically includes the following sub-steps: s1, input data: the ultrasonic signal characteristic parameters extracted in step 3 are used as input data and passed into the tensile strength prediction model; wherein the ultrasonic signal characteristic parameters include elastic modulus , wave speed , attenuation factor and frequency characteristics ; s2, feature extraction and processing: The convolutional neural network (CNN) is used to deeply extract time domain and frequency domain features, and the long short-term memory network (LSTM) is used to model the time series data of the signal, so as to further learn the elastic behavior of the pipeline and the dynamic changes of its tensile strength; and the weights of the neural network are updated by optimizing the back-propagation algorithm in this process; s3, tensile strength prediction model calculation: Use the trained CNN and LSTM models to analyze and calculate the input signal features to obtain the tensile strength prediction value of the pipeline. Calculated by the following formula:

[0023] in: It is the weight coefficient matrix obtained through historical data training, which is used for weight optimization and correction of prediction results; s4, weight coefficient adjustment: by comparing the error between historical experimental data and predicted values, the least square method or other appropriate regression algorithms are used to dynamically adjust the weight coefficients in the prediction model. To ensure the accuracy of the predicted results and their consistency with the actual tensile strength; s5, tensile strength output: output the tensile strength prediction value of the adjusted model It serves as an indicator of the current tensile performance of the pipeline and makes subsequent maintenance decisions.

[0024] As a further aspect of the present invention, the frequency range of the ultrasonic signal is 20 kHz to 300 kHz to increase the sensitivity to the pipe material and enhance the signal penetration.

[0025] As a further solution of the present invention, the ultrasonic transducer is a piezoelectric transducer, and the piezoelectric transducer is made of PZT lead titanate material, has high sensitivity and is adaptable to working requirements in high temperature environments.

[0026] As a further solution of the present invention, the signal processing module includes a time domain analysis unit and a frequency domain analysis unit. The time domain analysis unit is used to extract the time characteristics of the signal, and the frequency domain analysis unit is used to extract the frequency characteristics of the signal to further optimize the feature extraction of the signal.

[0027] As a further solution of the present invention, the deep learning model is a convolutional neural network CNN, and the CNN model includes at least three convolutional layers, a pooling layer and a fully connected layer to achieve deep learning of signal features and accurate prediction of tensile properties.

[0028] As a further solution of the present invention, the training data of the prediction model is derived from multivariate experimental data including different ambient temperatures, humidity, pipeline materials and manufacturing process parameters, and data consistency is ensured through data normalization processing.

[0029] As a further solution of the present invention, the deep learning algorithm is trained and optimized by the gradient descent method, and the cross-validation method is used to verify the generalization ability of the model to reduce overfitting.

[0030] As a further solution of the present invention, the tensile strength prediction result is combined with the actual use environment conditions of the pipeline, and risk assessment is performed by setting different thresholds. If the prediction result is lower than the safety threshold, maintenance suggestions are automatically generated and transmitted to relevant operators through the communication interface.

[0031] As a further solution of the present invention, the collection of ultrasonic reflection signals and transmission signals includes preprocessing the signals to remove background noise, and the noise removal method includes but is not limited to bandpass filter and wavelet transform to ensure high signal quality.

[0032] Embodiment 1: This embodiment provides an improved method for rapid detection of tensile properties of PE gas pipelines, so as to achieve accurate and non-destructive detection of pipeline tensile strength and adapt to a variety of environmental and material conditions.

[0033] Step 1: Installation of ultrasonic transducer and signal excitation.

[0034] First, at least one ultrasonic transducer is installed on the surface of the PE gas pipeline. The ultrasonic transducer uses a piezoelectric transducer or a laser transducer, and the operating frequency is set between 20kHz and 300kHz. This frequency range ensures that the ultrasonic wave can effectively penetrate the pipeline material and interact with the stress wave inside the pipeline, thereby providing the necessary signal characteristics for subsequent detection.

[0035] In this step, the ultrasonic transducer excites the ultrasonic guided wave to propagate along the axial direction of the pipeline and triggers stress waves in the pipeline, thereby exciting tiny vibrations on the pipeline surface. This operation is used to preliminarily obtain basic data on the pipeline status.

[0036] Step 2: Ultrasonic signal acquisition and feature extraction.

[0037] At least one ultrasonic receiving sensor is used to collect ultrasonic reflection and transmission signals after propagation through the pipeline. The propagation characteristics of the signal include parameters such as attenuation, frequency shift, propagation time, and reflection mode. These signal characteristics are the key to subsequent analysis and reflect the material properties, thickness, and internal defects of the pipeline.

[0038] The signal is processed finely through the time domain and frequency domain analysis units. The time domain analysis unit extracts the time characteristics of the signal, while the frequency domain analysis unit extracts the frequency characteristics. The combination of the two forms comprehensive characteristic data of the pipeline state, providing accurate candidate features for subsequent tensile strength prediction.

[0039] Step 3: Time-frequency analysis and signal processing of characteristic parameters.

[0040] The ultrasonic signal characteristic parameters (such as elastic modulus) extracted in step 2 , wave speed , attenuation factor , frequency characteristics ) to enter the signal processing module. In this step, wavelet transform or Fourier transform is used to perform time-frequency analysis on the signal. These methods can efficiently extract information such as elastic modulus and wave velocity of the pipeline and form characteristic data that can be used for tensile strength prediction.

[0041] Wavelet transform can provide high-resolution time-frequency domain analysis, which helps to capture rapidly changing signal characteristics; Fourier transform performs spectral analysis on the signal to further improve the understanding of pipeline status.

[0042] Step 4: Deep learning model training and prediction.

[0043] Based on the feature parameters extracted in step 3, a deep learning model is used to predict the tensile strength. This model combines a convolutional neural network (CNN) and a long short-term memory network (LSTM) to process time domain features and frequency domain features respectively. The convolutional neural network (CNN) includes at least three convolutional layers, a pooling layer, and a fully connected layer, which are used to extract the spatial features of the signal from the time domain and frequency domain. The convolutional neural network (CNN) is responsible for extracting the spatial features of the signal, and the LSTM network is used to analyze the time series data of the signal to accurately model the dynamic elastic behavior and tensile strength of the pipeline.

[0044] The deep learning model is trained using a large amount of historical experimental data, which includes factors such as different materials, pipe thicknesses, and environmental conditions (such as temperature and humidity), to ensure that the model can adapt to a variety of pipe specifications and environmental influences.

[0045] Step 5: Real-time prediction and tensile strength calculation.

[0046] After the model training is completed, the collected real-time signals are predicted through the trained CNN and LSTM models to calculate the tensile strength of the pipeline. The prediction result is obtained through the formula:

[0047] in, represents the predicted tensile strength, is the elastic modulus of the pipe, is the wave speed, is the attenuation factor, is the frequency characteristic, is the weight coefficient matrix obtained by training historical experimental data. It reflects the relationship between different pipeline parameters and tensile strength and is obtained through training optimization.

[0048] Step 6: Optimize and dynamically adjust the prediction results.

[0049] By comparing with historical experimental data, the prediction results are optimized using the error correction mechanism. The least squares method or other regression algorithms are used to dynamically adjust the prediction error and optimize the weight coefficient matrix. , to ensure the accuracy of the prediction results. The dynamic adjustment process includes a feedback mechanism, that is, by comparing the error between the predicted value and the actual tensile strength value, the prediction ability of the model is continuously corrected.

[0050] Step 7: Maintenance recommendation generation and risk assessment.

[0051] After the prediction results are completed, if the predicted tensile strength value is lower than the set safety threshold, the system will automatically generate maintenance suggestions. The maintenance suggestions are transmitted to the monitoring platform or on-site operators through the communication interface. In this way, timely measures can be taken to avoid possible risks of pipeline breakage or gas leakage, ensuring the safe operation of the pipeline system.

[0052] Example 3: In this example, we further refined the rapid detection method of the tensile properties of PE gas pipelines based on ultrasonic guided waves and deep learning algorithms. For example, in order to improve the quality of the signal and reduce noise interference, we used a combination of bandpass filters and wavelet transforms to pre-process the ultrasonic reflection signal and the transmission signal. Specifically, the bandpass filter first removes high-frequency and low-frequency noise to ensure that the core frequency components of the signal are effectively retained. Then, the wavelet transform is applied for multi-resolution analysis to further extract important features from the signal, especially in the analysis of non-stationary signals, and can effectively identify the characteristics of short-time transformations. This step ensures the high quality of the ultrasonic signal and reduces the impact of signal quality fluctuations in complex environments.

[0053] And to improve the accuracy and stability of the prediction model, the deep learning network uses a combination of a convolutional neural network (CNN) and a long short-term memory network (LSTM) in this embodiment. The convolutional neural network (CNN) includes at least three convolutional layers, a pooling layer, and a fully connected layer, which are used to extract the spatial features of the signal from the time domain and the frequency domain. The CNN part is responsible for extracting local features from the time domain and the frequency domain, while the LSTM part plays a role in time series data modeling, especially when dealing with the dynamic changes of the signal time series, it can accurately capture the tensile behavior of the pipeline. During the model training process, we normalized the data to ensure the consistency of the data under different experimental environments and reduce the influence of factors such as temperature, humidity, and pipeline material. Through the cross-validation method, we verified the generalization ability of the model and effectively reduced the overfitting phenomenon. And in the tensile strength prediction process, the least squares method is used to correct the predicted value by comparing it with the historical experimental data. Specifically, the least squares method dynamically adjusts the weight coefficient in the prediction model by calculating the error between the historical experimental data and the predicted value. This process optimizes the parameters of the neural network through the back propagation algorithm to ensure that the error between the accuracy of the prediction result and the actual tensile strength remains within the set tolerance range. Through this dynamic optimization mechanism, we can effectively cope with complex and changing environmental conditions and further improve the adaptability of the model.

[0054] In the construction site application scenario, multiple ultrasonic transducers can be fixedly installed on the surface of the PE pipe by means of a snap-on connection, for example, to excite ultrasonic guided waves. After the ultrasonic signal propagates through the pipe, it is collected by multiple ultrasonic receiving sensors. The signal processing module processes and analyzes the collected signals in real time, extracting parameters such as elastic modulus, wave velocity, attenuation factor, and frequency characteristics. The deep learning model analyzes these features through CNN and LSTM models, predicts the tensile strength of the pipeline in real time, and transmits the prediction results to the monitoring platform. Once the prediction result is lower than the set safety threshold, the system automatically generates maintenance suggestions and promptly notifies the on-site staff for follow-up processing, thereby avoiding the risk of pipeline breakage or leakage due to insufficient tensile strength. When training the deep learning model, experimental data from different environmental conditions and pipeline types were used, including temperature, humidity, strain rate, pipeline material and other factors. These data were strictly screened and standardized to ensure data consistency and comparability. With the support of large-scale experimental data, the model can provide reliable prediction results under different pipeline specifications and environmental conditions, and has strong generalization ability, which can adapt to the changing environment of the construction site.

[0055] Example 4: In this example, we further optimized the signal processing flow, model training method and weight coefficient adjustment mechanism, such as installing at least one ultrasonic transducer on the surface of the PE gas pipeline. The transducer is a piezoelectric transducer or a laser transducer, and the excitation frequency range is set to 20kHz to 300kHz. This frequency range ensures that the ultrasonic wave can effectively penetrate the pipeline and excite stress waves inside the pipeline. The transducer generates a tiny vibration on the surface of the pipeline through the excitation signal, and then obtains the information of the stress wave in the pipeline, providing necessary data support for subsequent signal acquisition and analysis.

[0056] After the ultrasonic signal propagates through the pipeline, at least one ultrasonic receiving sensor will receive the reflected signal and the transmitted signal. The propagation characteristics of these signals include attenuation, frequency shift, propagation time, etc., which are closely related to factors such as the material properties, thickness, and internal defects of the pipeline. Through time domain and frequency domain analysis, these signal features are extracted to form the pipeline status data. Specifically, the time domain analysis unit extracts the time characteristics of the signal, and the frequency domain analysis unit is responsible for extracting the frequency characteristics of the signal. For the extracted ultrasonic signal features, wavelet transform or Fourier transform is used to perform time-frequency analysis on the signal. Wavelet transform can provide high-resolution time-frequency domain analysis, especially suitable for capturing rapidly changing signal characteristics; while Fourier transform helps to perform spectral analysis of the signal and further optimize the understanding of the pipeline status. The goal of this process is to extract key pipeline characteristic parameters from the original signal, such as elastic modulus, wave velocity, attenuation factor, and frequency characteristics. After extracting the pipeline characteristic parameters, these features are further analyzed using deep learning algorithms (convolutional neural network CNN and long short-term memory network LSTM). CNN is responsible for processing local features in the time and frequency domains, while LSTM is specifically used to analyze the time series characteristics of the signal and capture the dynamic elastic behavior and changes in tensile strength of the pipeline. These models are trained using historical experimental data, which includes factors such as different materials, pipeline thickness, and environmental conditions (such as temperature and humidity), ensuring that the model can operate stably under a variety of pipeline specifications and environmental changes.

[0057] Then, the trained CNN and LSTM models are used to predict the tensile strength of the ultrasonic signals collected in real time. The prediction is based on the characteristic parameters output by the model to calculate the tensile strength of the pipeline. In order to ensure the accuracy of the prediction, the model will automatically perform error correction when the predicted value deviates from the true value. At this time, the error correction mechanism uses historical experimental data to compare with the current predicted value, and adjusts the weight coefficient in the model through a regression algorithm (such as the least squares method) to reduce the prediction error. This process dynamically optimizes the weight coefficient to ensure the accuracy and reliability of the tensile strength prediction. Once the prediction result is lower than the set safety threshold, the system will automatically generate maintenance recommendations. These recommendations will be transmitted to the on-site operators through the communication interface to guide them to inspect and maintain the pipeline. Through the risk assessment system, combined with the current environmental conditions and the predicted tensile strength value, the system can assess whether there is a potential risk of breakage or leakage in the pipeline. If the risk is too high, the system will provide relevant maintenance measures to ensure the safe operation of the pipeline system.

[0058] In the present invention, unless otherwise clearly stipulated and limited, the terms "installation", "setting", "connection", "fixation", "screw-on" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral one; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. Unless otherwise clearly defined, ordinary technicians in this field can understand the specific meanings of the above terms in the present invention according to the specific circumstances.

[0059] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present invention.

Claims

1. A rapid detection method for the tensile properties of PE gas pipelines, characterized by: The method comprises the following steps: Step 1, installing at least one ultrasonic transducer on the surface of the PE gas pipeline, for exciting ultrasonic guided waves to propagate along the axial direction of the pipeline and excite stress waves inside the pipeline, wherein the ultrasonic transducer is a piezoelectric transducer or a laser transducer, and the excitation frequency of the ultrasonic wave is between 10kHz and 500kHz, and the ultrasonic signal can propagate along the axial direction of the pipeline and interact with the stress waves inside the pipeline; Step 2, using at least one ultrasonic receiving sensor to collect ultrasonic reflection signals and transmission signals after propagation through the pipeline, and recording the propagation characteristics of the ultrasonic signal in real time, wherein the propagation characteristics of the ultrasonic signal include signal attenuation, frequency shift, propagation time and reflection mode parameters; Step 3, extracting features of the collected ultrasonic signal through the signal processing module, performing time-frequency analysis on the signal using wavelet transform or Fourier transform, and extracting the elastic modulus, wave velocity, attenuation factor and frequency characteristics of the pipeline as candidate characteristic parameters of the pipeline tensile strength; Step 4: Based on the extracted candidate feature parameters, a deep learning algorithm is used to analyze the ultrasonic signal through a convolutional neural network (CNN) and a long short-term memory network (LSTM) to construct a pipeline tensile strength prediction model, wherein the prediction model is obtained by training with historical experimental data, and the historical experimental data includes standard tensile strength data of PE pipelines under different materials, thicknesses, strain rates and environmental conditions; Step 5, using a prediction model to predict the ultrasonic signal in real time and calculate the tensile strength of the pipeline; Step 6: If the tensile strength prediction result is lower than the set threshold, a maintenance suggestion is automatically generated and transmitted to the monitoring platform or the on-site operator through the communication interface.

2. A rapid detection method for tensile properties of a PE gas pipeline according to claim 1, characterized in that: In step 5, the ultrasonic signal is predicted in real time by the prediction model to calculate the tensile strength of the pipeline, which specifically includes the following sub-steps: s1, input data: the ultrasonic signal characteristic parameters extracted in step 3 are used as input data and passed into the tensile strength prediction model; wherein the ultrasonic signal characteristic parameters include elastic modulus , wave speed , attenuation factor and frequency characteristics ; s2, feature extraction and processing: The convolutional neural network (CNN) is used to deeply extract time domain and frequency domain features, and the long short-term memory network (LSTM) is used to model the time series data of the signal, so as to further learn the elastic behavior of the pipeline and the dynamic changes of its tensile strength; and the weights of the neural network are updated by optimizing the back-propagation algorithm in this process; s3, tensile strength prediction model calculation: Use the trained CNN and LSTM models to analyze and calculate the input signal features to obtain the tensile strength prediction value of the pipeline. Calculated by the following formula: in: It is the weight coefficient matrix obtained through historical data training, which is used for weight optimization and correction of prediction results; s4, weight coefficient adjustment: by comparing the error between historical experimental data and predicted values, the least square method or other appropriate regression algorithms are used to dynamically adjust the weight coefficients in the prediction model. ; s5, tensile strength output: output the tensile strength prediction value of the adjusted model It serves as an indicator of the current tensile performance of the pipeline and makes subsequent maintenance decisions.

3. A rapid detection method for tensile properties of a PE gas pipeline according to claim 1, characterized in that: The frequency range of ultrasonic signals is 20kHz to 300kHz.

4. A rapid detection method for tensile properties of a PE gas pipeline according to claim 1, characterized in that: The ultrasonic transducer is a piezoelectric transducer, and the piezoelectric transducer is made of PZT lead titanate material.

5. A rapid detection method for tensile properties of a PE gas pipeline according to claim 1, characterized in that: The signal processing module includes a time domain analysis unit and a frequency domain analysis unit. The time domain analysis unit is used to extract the time characteristics of the signal, and the frequency domain analysis unit is used to extract the frequency characteristics of the signal.

6. A rapid detection method for tensile properties of a PE gas pipeline according to claim 1, characterized in that: The deep learning model is a convolutional neural network (CNN), and the CNN model includes at least three convolutional layers, a pooling layer, and a fully connected layer.

7. A rapid detection method for tensile properties of a PE gas pipeline according to claim 1, characterized in that: The training data of the prediction model comes from multivariate experimental data including different ambient temperatures, humidity, pipeline materials and their manufacturing process parameters, and data consistency is ensured through data normalization.

8. A rapid detection method for tensile properties of PE gas pipelines according to claim 1, characterized in that: The deep learning algorithm is trained and optimized through the gradient descent method, and the cross-validation method is used to verify the generalization ability of the model.

9. A rapid detection method for tensile properties of a PE gas pipeline according to claim 1, characterized in that: The tensile strength prediction results are combined with the actual environmental conditions of the pipeline, and risk assessment is performed by setting different thresholds. If the prediction result is lower than the safety threshold, maintenance recommendations are automatically generated and transmitted to relevant operators through the communication interface.

10. A rapid detection method for tensile properties of a PE gas pipeline according to claim 1, characterized in that: The collection of ultrasonic reflection signals and transmission signals includes preprocessing the signals to remove background noise, and the noise removal method includes but is not limited to bandpass filter and wavelet transform.

Images