CN-122020327-A - Method and device for predicting corrosion defect of small-caliber pipeline and electronic equipment

CN122020327ACN 122020327 ACN122020327 ACN 122020327ACN-122020327-A

Abstract

The invention belongs to the technical field of detection based on machine learning, and provides a method, a device and electronic equipment for predicting corrosion defects of a small-caliber pipeline, wherein the method comprises the steps of constructing a corrosion depth quantization model of a support vector machine based on a training set and a testing set; the method comprises the steps of obtaining an optimization algorithm adopted for optimizing a corrosion depth quantization model, obtaining optimization parameters of the optimization algorithm, searching an optimal solution in a parameter space of a support vector machine based on the optimization parameters, optimizing the corrosion depth quantization model based on the optimal solution to obtain the corrosion depth quantization optimization model, inputting a plurality of test samples in a test set into the corrosion depth quantization optimization model, predicting corrosion defects of a target small-caliber pipeline, and obtaining and outputting a prediction result. By the prediction method, the corrosion defect depth of the target small-caliber pipeline can be quantized, so that the corrosion degree of the target small-caliber pipeline can be accurately determined based on the corrosion defect depth quantized value.

Inventors

LIU WENCAI
GAO ZHIJIE
JIANG RUIJING
LI YINGLI
KANG LE
QIAO WEI
ZHENG HUIYONG
Sun Jipei

Assignees

中国石油天然气股份有限公司
中国石油集团安全环保技术研究院有限公司

Dates

Publication Date: 20260512
Application Date: 20241112

Claims (10)

1. A method for predicting corrosion defects in a small-bore pipeline, the method comprising: acquiring an experimental data set acquired through a detection platform, and dividing the experimental data set into a training set and a testing set, wherein the detection platform is used for detecting corrosion defects of a target small-caliber pipeline; Generating sinusoidal excitation signals by using a signal generator, collecting lamb wave signals at a plurality of depths, and extracting time domain features of the lamb wave signals at the plurality of depths to obtain time domain feature data; Constructing a corrosion depth quantization model of a support vector machine based on the training set and the test set; acquiring an optimization algorithm adopted for optimizing the corrosion depth quantization model, and acquiring optimization parameters of the optimization algorithm; searching an optimal solution in a parameter space of a support vector machine based on the optimization parameters, and optimizing the corrosion depth quantization model based on the optimal solution to obtain a corrosion depth quantization optimization model; And inputting a plurality of test samples in the test set into the corrosion depth quantification optimization model, predicting the corrosion defect of the target small-caliber pipeline to obtain and output a prediction result, wherein the prediction result comprises corrosion defect depth quantified values of the plurality of test samples and corrosion defect depths of the plurality of test samples under model quantification.
2. The method for predicting according to claim 1, wherein, The optimization algorithm is a corrosion depth quantization algorithm of the support vector machine, which is obtained by optimizing a firefly algorithm, and the optimization parameter is an optimization parameter for corrosion depth quantization of the support vector machine.
3. The prediction method according to claim 2, wherein the obtaining the optimization parameters of the optimization algorithm includes: Acquiring a parameter upper bound and a parameter lower bound, and randomly initializing optimized parameters between the parameter upper bound and the parameter lower bound; iteratively updating the light absorption coefficient, and calculating the mutual attraction degree between any two fireflies in the fireflies; comparing brightness between any two fireflies in the plurality of fireflies, and updating firefly positions; According to the updated firefly position of any firefly, recalculating the brightness of the firefly to obtain updated firefly brightness; Judging whether the current condition meets the iteration stop condition or not based on the updated firefly brightness of any firefly, obtaining and outputting the optimized parameter if the current condition is judged to meet the iteration stop condition, otherwise, continuing to iteratively update the light absorption coefficient until the current condition meets the iteration stop condition, and obtaining and outputting the optimized parameter.
4. The prediction method according to claim 1, wherein after the obtaining and outputting of the prediction result, the method further comprises: acquiring real values of corrosion defect depths sequentially corresponding to a plurality of test samples, and acquiring actual depths of corrosion defects sequentially corresponding to the plurality of test samples; Comparing the real corrosion defect depth values sequentially corresponding to the plurality of test samples with the quantized corrosion defect depth values sequentially corresponding to the plurality of test samples to verify the accuracy of the prediction result based on the corresponding comparison result; comparing the actual depth of the corrosion defect sequentially corresponding to the plurality of test samples with the depth of the corrosion defect in the model quantification sequentially corresponding to the plurality of test samples, so as to verify the accuracy of the prediction result based on the corresponding comparison result.
5. The prediction method according to claim 1, wherein the time domain feature extraction of the time domain features of the lamb wave signals at a plurality of depths includes: Acquiring a plurality of characteristic parameters, wherein the plurality of characteristic parameters comprise a plurality of dimensional parameters and a plurality of dimensionless parameters; Obtaining the plurality of dimensional parameters and a corresponding first extraction formula, wherein the plurality of dimensional parameters comprise peak-to-peak values, average amplitude values, variances, standard deviations, kurtosis, skewness, root mean square, mean square values and square root amplitude values; performing time domain feature extraction on any one of the plurality of dimensional parameters through a corresponding first extraction formula; Obtaining the plurality of dimensionless parameters and a corresponding second extraction formula, wherein the plurality of dimensionless parameters comprise kurtosis factors, skewness coefficients, margin factors, waveform coefficients, peak coefficients, pulse factors and spearman correlation coefficients; And carrying out time domain feature extraction on any one of the plurality of dimensionless parameters through a corresponding second extraction formula.
6. The method of predicting as recited in claim 5, further comprising: Analyzing the correlation between any one of the characteristic parameters and the defect depth of the corrosion defect to obtain a correlation analysis result; And analyzing the trend between any one of the characteristic parameters and the defect depth of the corrosion defect to obtain a trend analysis result.
7. The method of claim 1, wherein prior to said acquiring the experimental data set acquired by the detection platform, the method further comprises: the detection platform is built through a plurality of devices, and the plurality of devices comprise: A plurality of excitation sensors and a plurality of receiving sensors arranged on the circumferential surface of the target small-caliber pipeline, wherein the plurality of receiving sensors are sequentially and equally angularly distributed on the circumferential surface at equal intervals; the multiple experimental instruments comprise piezoelectric ceramic columns, acoustic emission sensors, an arbitrary function signal generator, a multi-channel direct-current stabilized voltage supply, a high-speed data acquisition card, an upper computer and an ultrasonic thickness gauge.
8. A device for predicting corrosion defects in small-bore pipes, the device comprising: The first acquisition module is used for acquiring an experimental data set acquired through a detection platform and dividing the experimental data set into a training set and a testing set, and the detection platform is used for detecting corrosion defects of a target small-caliber pipeline; The extraction module is used for generating a sine excitation signal by using the signal generator, collecting lamb wave signals at a plurality of depths, and extracting time domain features of the lamb wave signals at the plurality of depths to obtain time domain feature data; The model construction module is used for constructing a corrosion depth quantization model of the support vector machine based on the training set and the testing set; The second acquisition module is used for acquiring an optimization algorithm adopted for optimizing the corrosion depth quantization model and acquiring optimization parameters of the optimization algorithm; the model optimization module is used for searching an optimal solution in a parameter space of the support vector machine based on the optimization parameters, and optimizing the corrosion depth quantization model based on the optimal solution to obtain a corrosion depth quantization optimization model; The prediction module is used for inputting a plurality of test samples in the test set into the corrosion depth quantization optimization model, predicting the corrosion defect of the target small-caliber pipeline, and obtaining and outputting a prediction result, wherein the prediction result comprises corrosion defect depth quantized values sequentially corresponding to the plurality of test samples and corrosion defect depths in model quantization sequentially corresponding to the plurality of test samples.
9. A computer readable storage medium, characterized in that a computer program is stored thereon, which, when executed in a computer, causes the computer to perform the method of any of claims 1 to 7.
10. An electronic device comprising a memory and a processor, the memory having executable code stored therein, the processor, when executing the executable code, implementing the method of any one of claims 1 to 7.

Description

Method and device for predicting corrosion defect of small-caliber pipeline and electronic equipment Technical Field The disclosure belongs to the technical field of detection based on machine learning, and particularly relates to a method and a device for predicting corrosion defects of a small-caliber pipeline and electronic equipment. Background Pipelines are one of the economic means for transporting the medium such as petroleum, natural gas and the like, and occupy important positions in national economy and production. However, the pipe may also fail due to corrosion or the like. The large amount of medium and small caliber oil gas conveying pipelines in our country are distributed in areas with denser population, and once corrosion leakage accidents occur, the consequences are more serious. Therefore, detection of oil and gas pipelines is particularly important. In the petrochemical industry, small-caliber pipelines are used in various aspects of petrochemical pipeline systems, and the application comprises 1) pipeline transportation, 2) measurement and instrument, real-time monitoring and control of process parameters by installing measurement instruments such as flow meters, pressure sensors and temperature sensors on the small-caliber pipelines, and ensuring the safety and quality of production processes, 3) process control, wherein the small-caliber pipelines are used in a process control system of a petrochemical plant and are connected with control valves, regulating valves and other control elements, and 4) equipment connection and installation, wherein the small-caliber pipelines are used as connection elements for connecting various equipment and machinery in the petrochemical plant, such as pumps, compressors, heat exchangers and the like. Although the small-bore pipeline is generally made of high-strength materials with corrosion resistance and high temperature resistance, due to the severe working environment (in humid, corrosive gas or water and vibration load), the pipeline system in long-term operation inevitably has defects such as oxidation, corrosion, metal fatigue, cracks and the like. The detection is only carried out qualitatively, and the accuracy of the quantitative characterization detection of the defects of the small-caliber pipeline is a challenge which needs to be overcome, and is also a difficult problem to be solved in the petrochemical industry. In the early-stage storage tank detection process, the traditional detection methods such as ultrasonic, magnetic powder, rays and permeation are generally adopted, but the surface roughness or unevenness of a small-caliber pipeline can interfere with the transmission of ultrasonic waves, so that the detection result is inaccurate, the magnetic powder detection is only suitable for ferromagnetic materials, the applicability is poor for non-ferromagnetic materials such as stainless steel or aluminum alloy small-caliber pipelines, the ray detection process is slower, the cost is higher, the large-scale industrial production is not utilized, the permeation detection requires the detected surface to be clean, and the detection method is difficult to be applied to the storage tank in operation. The ultrasonic guided wave detection detects defects through the transmission and reflection characteristics of guided waves, has the characteristics of high efficiency and high speed, and is suitable for pipeline detection. The ultrasonic guided wave is used for quantitatively characterizing the defects of the small-caliber pipeline, and has the following difficulties that 1) the complexity of the pipeline structure can cause attenuation and dispersion of the guided wave, the sensor receives the deformation of the guided wave signal and has serious influence on quantitative identification of the depth of the corrosion defect, 2) the corrosion depth does not have a good correlation with the characteristic parameters extracted from the depth of the corrosion defect under the Lamb wave transmission sound field, 3) the accuracy of a corrosion depth quantitative model is low, the error is large, 4) the defects are not directly classified, secondary detection is needed, and time and labor are consumed. How to provide a method capable of accurately predicting corrosion defects of a small-caliber pipeline is a technical problem to be solved. Disclosure of Invention Based on the above, it is necessary to provide a method, an apparatus, a storage medium and an electronic device for predicting the corrosion defect of a small-caliber pipeline, aiming at the defect that the prior art cannot accurately predict the corrosion defect of the small-caliber pipeline. In a first aspect, an embodiment of the present invention provides a method for predicting corrosion defects of a small-caliber pipe, the method including: acquiring an experimental data set acquired through a detection platform, and dividing the experimental data set into a training set and a testi