Pipeline cracking damage sonar intelligent evaluation method
By extracting the correlation and energy characteristic parameters of sonar echo signals, and combining the GMM model and Frechet similarity assessment, the problem of accurate assessment of cracking damage in complex urban underground pipeline environments was solved, achieving more efficient damage state monitoring and probabilistic assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-04-25
- Publication Date
- 2026-04-07
AI Technical Summary
In urban underground pipelines, existing technologies are insufficient to accurately assess cracking damage in complex and harsh environments. Single GMM models are susceptible to environmental interference, resulting in large assessment errors and making it impossible to achieve real-time monitoring and probabilistic assessment.
Correlation and energy characteristic parameters of sonar echo signals are extracted, parameter distribution is analyzed using the GMM method, a benchmark GMM library is constructed, and damage status is assessed through Frechet similarity to overcome environmental interference and achieve probabilistic assessment.
It improves the accuracy and scientific rigor of damage assessment in complex environments, has better resistance to environmental interference, and with the abundance of monitoring data, the assessment results are more reasonable and accurate.
Smart Images
Figure CN116465977B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the field of urban underground pipeline damage sonar evaluation, and relates to a pipeline cracking damage sonar intelligent evaluation method. BACKGROUND
[0002] The pipeline is an important component of urban operation, and undertakes important tasks in rainwater discharge, sewage discharge, domestic water supply, energy supply, etc. The pipeline is prone to cracking damage under long-term use, which is difficult to be directly detected and is easy to cause pipeline burst, sewage leakage and other accidents. Therefore, detecting and evaluating the cracking state of the pipeline is very important for ensuring the stable operation of the infrastructure.
[0003] Sonar is a commonly used liquid-filled pipeline damage detection technology, which can detect the cracking degree of the pipeline and realize damage evaluation by using the echo attenuation law of the sonar. Although the echo amplitude will change with the size of the crack, in the underground harsh working condition environment, water flow, impurities, pipe wall holes and other factors will affect the change of the sonar echo, which leads to errors in evaluating the damage state by using the echo signal attenuation, and is not conducive to real-time monitoring of the pipeline damage state.
[0004] The pipeline cracking damage is a dynamic and gradual process, and the massive data information accumulated in continuous monitoring can provide cracking damage development information. Combined with intelligent algorithms to mine this information, the probabilistic evaluation of cracking damage can be realized. For example, the patent application CN114811452A uses a GMM model to predict pipeline leakage points, which can realize the evaluation of buried pipeline damage position. However, there are many scenarios such as full liquid, half liquid, and solid-liquid mixture in the internal medium of urban underground pipelines. In such complex working condition environment, a single GMM model will be disturbed by the environment, and it is difficult to accurately evaluate the difference between the monitoring working condition and the reference working condition. Therefore, the single GMM model method is difficult to apply in the complex and harsh scene of urban underground pipelines. At present, there is no good theoretical system to apply this idea to realize the probabilistic and intelligent evaluation of pipeline cracking damage by the correlation of GMM distribution between the reference working condition and the monitoring working condition. SUMMARY
[0005] To solve the above problems, the application provides a pipeline cracking damage sonar intelligent evaluation method, which extracts two characteristic parameters of sonar echo signal correlation and energy, and then analyzes the parameter distribution by using a GMM method to evaluate the pipeline state.
[0006] The pipeline cracking damage sonar intelligent evaluation method provided by the application comprises the following steps:
[0007] Step 1, obtaining a sonar echo signal, extracting two characteristic parameters of signal correlation and energy;
[0008] Step 2, analyze the distribution characteristics of parameters under different damage states by using GMM (Gaussian Mixture Model), and construct a GMM reference library of known damage states;
[0009] Step 3, for the sonar signal of unknown damage state, extract signal correlation and energy characteristic parameters, and calculate the GMM probability density distribution;
[0010] Step 4, evaluate the Frechet similarity of the damage state signal and the GMM reference library, and determine which damage state it belongs to.
[0011] Further, in step 1, two characteristic parameters of signal correlation and energy are extracted, and the specific steps are as follows:
[0012] (1) Take the undamaged state as the reference signal, and calculate the correlation parameters of the sonar echo signal under different damage states and the reference signal:
[0013]
[0014] (2) The energy characteristic parameters of different working condition sonar echo signals are:
[0015]
[0016] Where i is the number of working conditions, f i (t) is the damage working condition signal, f s (t) is the undamaged state signal, t p , t q are the starting time and ending time of the effective signal respectively.
[0017] Further, in step 2, the GMM reference library of known damage states is constructed, and the specific steps are as follows: for the sonar signal of known damage state, the distribution characteristics of signal correlation parameters and energy characteristic parameters are analyzed by using GMM, and the probability density distribution function is:
[0018]
[0019] Where D k， k=1,2 are the characteristic parameters to be analyzed; μ is the mean of the probability density distribution, ∑ is the covariance of the probability density distribution; μ i is the mean of the probability density distribution of each component; ∑ i is the covariance of the probability density distribution of each component; P is the number of Gaussian components; ω i is the weight of each component; N i is the probability density function of each Gaussian component, and the expression is:
[0020]
[0021] Q represents the parameter dimension; the EM algorithm is used for optimization during parameter iteration, and the iterative formulas for weights, mean, and covariance are as follows:
[0022]
[0023]
[0024]
[0025] in Let E(γ(D)|ω) represent the weight, mean, and covariance of the p-th component in the i-th iteration, respectively, where K is the number of data points involved in the calculation. i ,μ i ,∑ i H(D) is the estimation function for the classification of the sample. The obtained GMM probability density distribution is used as the baseline data for this damage state, denoted as H(D). p The same operation is applied to all damage states to construct a GMM benchmark library.
[0026] Furthermore, in step 3, two feature parameters, signal correlation and energy, are extracted from the sonar signal with unknown damage state. Then, using these two feature parameters as input, the GMM probability density distribution of this signal is calculated, and the result is denoted as H(D). q ).
[0027] Furthermore, in step 3, the Frechet similarity between the damage state signal and the GMM benchmark library is evaluated. Specifically, the Frechet distance is used to characterize the similarity between the benchmark state and the damage state. The Frechet similarity between the benchmark state and the damage state is as follows:
[0028]
[0029] in The distance is Euclidean. The signal is classified as having the highest similarity to a certain state under a reference condition.
[0030] The beneficial effects of this invention are as follows: Considering the complex and variable internal media of urban underground pipelines and the harsh engineering environment, this invention utilizes the Gaussian Mixture Model (GMM) to analyze the parameter distribution under different damage states and constructs a benchmark GMM library. In testing applications, the correlation between the monitored state and the benchmark state is assessed through the Frechet similarity between the signal and the benchmark library. This overcomes the errors caused by harsh environmental interference in a single GMM model from a probabilistic perspective, making the evaluation results more reasonable. This invention uses a data-driven probabilistic method to evaluate damage states, possessing better resistance to environmental interference. As monitoring data becomes increasingly abundant, the evaluation results become increasingly accurate and scientific. Attached Figure Description
[0031] Figure 1 is an acoustic echo signal diagram of different damage states;
[0032] Figure 2 is a reference GMM distribution diagram of training signal construction;
[0033] Figure 3 is a GMM distribution diagram of application stage signal;
[0034] Figure 4 is a Frechet similarity diagram.
[0035] Figure 5 is a flow chart of the method of the present application. DETAILED DESCRIPTION
[0036] In order to make the content of the present application more easily and clearly understood, the present application will be further described in detail below according to specific embodiments and in conjunction with the accompanying drawings.
[0037] As shown in Figure 5 , a pipeline cracking damage acoustic intelligent evaluation method of the present application, the steps are:
[0038] Step 1, obtain the acoustic echo signal, extract the signal correlation and energy two characteristic parameters;
[0039] Step 2, use GMM (Gaussian Mixture Model) to analyze the distribution characteristics of the parameters under different damage states, and construct a GMM reference library of known damage states;
[0040] Step 3, for the acoustic signal of unknown damage state, extract the signal correlation and energy characteristic parameters, and calculate the GMM probability distribution;
[0041] Step 4, evaluate the Frechet similarity of the damage state signal and the GMM reference library, and determine which damage state it belongs to.
[0042] Among them, the signal correlation and energy two characteristic parameters in step 1 are extracted, and the specific steps are:
[0043] (1) Take the undamaged state as the reference signal, calculate the correlation parameters of the acoustic echo signal under different damage states and the reference signal:
[0044]
[0045] (2) The energy characteristic parameters of the acoustic echo signal under different working conditions are:
[0046]
[0047] Where i is the number of working conditions, f i(t) is the damage condition state signal, f s (t) is the undamaged state signal, t p , t q are the start time and end time of the effective signal respectively.
[0048] In step 2, the GMM reference library of known damage states is constructed, and the specific steps are as follows: for the sonar signal of the known damage state, the distribution characteristics of the signal correlation parameter and the energy characteristic parameter are analyzed by using the GMM, and the probability density distribution function is:
[0049]
[0050] Where D k is the characteristic parameter to be analyzed, k=1, 2 in the present application; μ is the mean of the probability density distribution, and ∑ is the covariance of the probability density distribution; μ i is the mean of the probability density distribution of each component; ∑ i is the covariance of the probability density distribution of each component; P is the number of Gaussian components; ω i is the weight of each component; N i is the probability density function of each Gaussian component, and the expression is:
[0051]
[0052] Where Q is the parameter dimension; in the parameter iteration, the EM algorithm is used for optimization, and the iterative formula of the weight, mean and covariance is:
[0053]
[0054]
[0055]
[0056] Where represent the weight, mean and covariance of the p-th component in the i-th iteration respectively, K is the number of data points participating in the operation, and E(γ(D)|ω i , μ i , ∑ i ) is the estimation function of the sample classification. The obtained GMM probability density distribution is taken as the reference data of this damage state, denoted as H(D p ); the same operation is performed on all damage states to construct the GMM reference library.
[0057] In step 3, for the sonar signal of the unknown damage state, two characteristic parameters of signal correlation and energy are extracted. Then, the GMM probability distribution of this signal is calculated by taking the two characteristic parameters as inputs, and the result is denoted as H(D q );
[0058] The Frechet similarity of the damage state signal and the GMM reference library is evaluated, and the specific method is as follows: the Frechet distance is used to represent the similarity of the reference state and the damage state, and the Frechet similarity of the reference state and the damage state is:
[0059]
[0060] Wherein is the Euclidean distance. When the sonar echo is most similar to a state in the reference state, the signal is classified as this kind of damage state.
[0061] The crack length is set to 1-4 levels to simulate four kinds of cracking damage states. The sonar echoes reflected by pipes with different crack sizes are as shown in Figure 1 , and the figure shows that when the damage is small, the sonar echo has two obvious wave peaks, and as the cracking increases, the proportion of single-wave peak echo signal also increases. Two characteristic parameters of each signal are extracted and standardized, the distribution is analyzed by using GMM, the reference GMM library is constructed, and the result is as shown in Figure 2 . In the test application stage, one sample of damage state 1-4 is selected, the parameters are extracted and the GMM distribution is calculated, and the result is as shown in Figure 3 . The Frechet similarity with the reference GMM is calculated, and the result is as shown in Figure 4 , and the figure shows that the maximum similarity of the damage state 1-4 sample is consistent with the state in the reference GMM library, and the similarity is greater than 90%, which shows that this kind of probabilistic damage evaluation method is feasible.
[0062] The above only describes the preferred scheme of the present application, and is not as a further limitation of the present application, and any equivalent changes made by using the content of the present application and the drawings are within the protection scope of the present application.
Claims
1. A sonar-based intelligent assessment method for pipeline cracking damage, characterized in that, The method includes: Step 1: Acquire sonar echo signals and extract two characteristic parameters: signal correlation and energy. Step 2: Analyze the distribution characteristics of parameters under different damage states using GMM, and construct a GMM benchmark library for known damage states; Step 3: Extract signal correlation and energy characteristic parameters from sonar signals with unknown damage status, and calculate the GMM probability density distribution; Step 4: Evaluate the similarity between the damage state signal and the Frechet data in the GMM benchmark library to determine which damage state it belongs to; In step 1, two feature parameters, signal correlation and energy, are extracted. The specific steps are as follows: (1) Using the undamaged state as the reference signal, calculate the correlation parameters between the sonar echo signal and the reference signal under different damage states: ； (2) The energy characteristic parameters of sonar echo signals under different operating conditions are as follows: ； Where i is the number of working condition types, f i (t) represents the damage condition signal, f s (t) represents the undamaged state signal, t p t q These are the start and end times of the valid signal, respectively. Step 2 involves constructing a GMM benchmark library with known damage states. The specific steps are as follows: For sonar signals with known damage states, the distribution characteristics of signal correlation parameters and energy characteristic parameters are analyzed using GMM. The probability density distribution function is as follows: ； Where D k k=1,2 are the characteristic parameters to be analyzed; Let be the mean of the probability density distribution. Let be the covariance of the probability density distribution; This represents the mean of the probability density distribution of each component; Let be the covariance of the probability density distribution of each component; P is the number of Gaussian components. The weights of each component; N i The probability density function for each Gaussian component is expressed as: ； Q represents the parameter dimension; the EM algorithm is used for optimization during parameter iteration, and the iterative formulas for weights, mean, and covariance are as follows: ， ， ， in , , Let represent the weight, mean, and covariance of the p-th component in the i-th iteration, respectively, and K be the number of data points involved in the calculation. The estimated function for sample classification; the obtained GMM probability density distribution is used as the baseline data for this damage state, denoted as H(D). p The same operation is applied to all damage states to construct a GMM benchmark library.
2. The intelligent sonar assessment method for pipeline cracking damage according to claim 1, characterized in that, In step 3, for sonar signals with unknown damage states, two feature parameters, signal correlation and energy, are extracted. Then, using these two feature parameters as input, the GMM probability density distribution of this signal is calculated, and the result is denoted as H(D). q ).
3. The intelligent sonar assessment method for pipeline cracking damage according to claim 1, characterized in that, In step 4, the Frechet similarity between the damage state signal and the GMM benchmark library is evaluated. Specifically, the Frechet distance is used to characterize the similarity between the benchmark state and the damage state. The Frechet similarity between the benchmark state and the damage state is as follows: ； in The distance is Euclidean; when the sonar echo has the highest similarity to a certain state under the reference state, the signal is classified as this damage state.