CN114154539A - High-sulfur-content gas well casing defect identification method based on direct-current magnetic field and integrated learning - Google Patents

High-sulfur-content gas well casing defect identification method based on direct-current magnetic field and integrated learning Download PDF

Info

Publication number
CN114154539A
CN114154539A CN202111435387.XA CN202111435387A CN114154539A CN 114154539 A CN114154539 A CN 114154539A CN 202111435387 A CN202111435387 A CN 202111435387A CN 114154539 A CN114154539 A CN 114154539A
Authority
CN
China
Prior art keywords
sulfur
gas well
defect
magnetic field
casing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111435387.XA
Other languages
Chinese (zh)
Inventor
黄平捷
任昊
侯迪波
赵腾
王晓伟
上官培俊
喻洁
张光新
张宏建
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN202111435387.XA priority Critical patent/CN114154539A/en
Publication of CN114154539A publication Critical patent/CN114154539A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Software Systems (AREA)
  • Medical Informatics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Signal Processing (AREA)
  • Probability & Statistics with Applications (AREA)
  • Investigating Or Analyzing Materials By The Use Of Magnetic Means (AREA)

Abstract

The invention discloses a method for identifying casing defects of a high-sulfur-content gas well based on a direct-current magnetic field and ensemble learning. Firstly, preprocessing data by using a moving average method to realize the phase alignment of direct current magnetic field signals under the same defect type; then, performing feature extraction on the direct current magnetic field signal by using principal component analysis; and finally, combining a plurality of SVM classifiers with the classification result of the decision tree in a weighted average mode by utilizing a bagging algorithm of integrated learning to realize the identification of various casing defects. The method for identifying the defects of the high-sulfur-content gas well casing based on the direct-current magnetic field and the integrated learning effectively extracts the distortion characteristics of the casing, has high classification precision, and has good generalization performance on different well conditions of the high-sulfur-content gas well, so that the method can be effectively applied to casing deformation condition monitoring of the high-sulfur-content gas well.

Description

High-sulfur-content gas well casing defect identification method based on direct-current magnetic field and integrated learning
Technical Field
The invention belongs to the field of nondestructive testing of conductive structures, and particularly relates to a method for identifying defects of a high-sulfur-content gas well casing based on a direct-current magnetic field and integrated learning.
Background
In the modern oil and gas industry, a sleeve is required to be added on the outer layer of an oil pipe for oil and gas collection to support a well wall and maintain normal operation of an oil and gas well. However, for high sulfur-containing gas wells, the working conditions are severe, and the problem that the casing deforms under external pressure in the salt-gypsum rock stratum section is more prominent, including: single-side extrusion, bending, transverse and longitudinal seams and the like affect the safety of oil and gas delivery in the oil pipe. Knowing the type of the casing defects can help to know the underground geological conditions and make corresponding repair measures and oil and gas collection measures. The method is limited by an integrated permanent well completion pipe column of the high-sulfur gas well, the deformation condition of the casing can not be measured through an oil pipe by the existing conventional multi-arm hole diameter, underground televisions and other well logging means at home and abroad, the methods such as pulse eddy current detection, infrasonic wave detection and the like are mostly suitable for single-pipe detection, and large errors exist in double-pipe detection. The direct current magnetic field signal has a good effect on double-pipe detection, so that how to judge the defect type of the casing based on the measured direct current magnetic field data has important significance on exploitation of the high-sulfur-content gas well.
At present, few researches on the direct-current magnetic field signal-based casing defect classification method are performed at home and abroad, and more mature researches are mainly performed on the defect classification method based on the pulse eddy current signal and the magnetic leakage signal. For example, golden aviation and the like combine the time domain characteristics and the frequency domain characteristics of the response signals with the pulse eddy current signals, and classify the casing defects by using an SVM classifier. Zhangyu et al propose a classification method based on stacked self-coding neural networks. When the classification methods are directly applied to the direct-current magnetic field signals, the defect distortion characteristics in the direct-current magnetic field signals are difficult to effectively extract, and the classification accuracy is poor.
Disclosure of Invention
The invention aims to provide a method for identifying the defects of a high-sulfur-content gas well casing based on a direct-current magnetic field and integrated learning, aiming at the defects of the prior art. The invention adopts the Bagging algorithm of integrated learning to combine a plurality of weak classifiers.
The purpose of the invention is realized by the following technical scheme: a high-sulfur-content gas well casing defect identification method based on a direct-current magnetic field and integrated learning comprises the following steps:
s1: and scanning and detecting the defective casing of the high-sulfur-content gas well by using a direct-current magnetic field sensor to obtain a static magnetic signal of the casing defect, and marking according to the actual defect type.
S2: the static magnetic signal acquired in step S1 is preprocessed.
S3: and (4) performing data shift on the static magnetic signal preprocessed in the step (S2) by using a moving average method, and realizing the phase alignment of the distorted parts under the same defect type.
S4: the data phase-aligned in step S3 is subjected to feature extraction by principal component analysis.
S5: based on the features extracted in step S5 and their corresponding defect types, a multi-classifier is trained. And combining the classification results of the weak classifiers in a weighted average mode by utilizing a bagging algorithm of ensemble learning.
S6: and (4) processing the high-sulfur-content gas well casing to be tested in the steps S1-S4, extracting corresponding characteristics, inputting the characteristics into the multi-classifier trained in the step S5, and obtaining a model-predicted high-sulfur-content gas well casing defect identification result.
Further, in step S1, the defect types include: single-sided extrusion, double-sided extrusion, bending, transverse seam, longitudinal seam, hole, four-sided extrusion and no defect.
Further, in step S3, the static magnetic signal data is cyclically shifted, the pearson correlation coefficients for different shifts are calculated, and the shift length at the minimum pearson correlation coefficient is set as the optimum shift length, so that the static magnetic signal after data shift is obtained.
Further, in step S2, the preprocessing includes smoothing filtering, data centralization, and the like.
Further, in step S5, the weak classifier is an SVM classifier or a decision tree classifier.
Further, in step S5, the multiple classifiers include 2n weak classifiers, where the weak classifiers are binary models, specifically n SVM classifiers and n decision trees, and n is the number of defect types. Each weak classifier predicts the probability of a particular defect type and other defect types, respectively, and each defect type is predicted by an SVM classifier and a decision tree classifier, respectively. And performing weighted summation on the probability of each defect type based on the prediction probability of the weak classifier, and finally performing normalization through a softmax function to output the probability of each defect type, wherein the output with the highest probability is the defect type.
The invention has the beneficial effects that: the method effectively extracts the distortion characteristics of the casing, has higher classification precision, and has good generalization performance for different well conditions of the high-sulfur-content gas well, so that the method can be effectively applied to casing deformation condition monitoring of the high-sulfur-content gas well.
Drawings
FIG. 1 is a flow chart of the present invention for identifying casing defects in a high sulfur gas well based on DC magnetic field and ensemble learning;
FIG. 2 is a schematic diagram of the ensemble learning algorithm of the present invention.
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings:
as shown in fig. 1, the method for identifying the casing defect of the high-sulfur-content gas well based on the direct-current magnetic field and the integrated learning of the invention comprises the following steps:
s1: and scanning and detecting the defective casing of the high-sulfur-content gas well by using a direct-current magnetic field sensor to obtain a static magnetic signal of the casing defect, and marking the corresponding actual defect type. The defect types include: single-sided extrusion, double-sided extrusion, bending, transverse seam, longitudinal seam, hole, four-sided extrusion and no defect.
S2: the static magnetic signal acquired in step S1 is subjected to preprocessing such as smoothing filtering and data centering.
S3: and circularly shifting the static magnetic signal data preprocessed in the step S2 by using a moving average method, obtaining a pearson coefficient between the static magnetic signals under different shifting conditions, and performing data shifting on the static magnetic signals by taking a shift length under the minimum pearson correlation coefficient as an optimal shift length, thereby realizing phase alignment of measurement signals (distorted parts) at different depths under the same defect type.
S4: and dividing the data after the phase alignment in the step S3 into a training set and a test set, wherein the dividing method is a Bootstrapping method.
S5: the feature extraction is performed on the data after the phase alignment in step S3 by using principal component analysis.
S6: and training the multi-classifier based on the training set features extracted in the step S5 and the corresponding defect types. And combining the classification results of the weak classifiers in a weighted average mode by using a Bagging algorithm of ensemble learning.
As shown in fig. 2, the multi-classifier includes 2n weak classifiers, which are binary models, specifically n SVMs and n decision trees, where n is 8, which is the number of defect types to be classified. Each weak classifier predicts the probability of a certain defect type and other defect types respectively, and each defect type is predicted twice by an SVM classifier and a decision tree respectively. And the probability of each defect type is weighted and summed based on the prediction probabilities of all weak classifiers, finally, the probability of each defect type is normalized through a softmax function, the probability of each defect type is finally output by the multi-classifier, and the output with the highest probability is the defect type.
S7: and (4) inputting the test set characteristics extracted in the step S5 into the multi-classification model trained in the step S6 to obtain the high-sulfur-content gas well casing defect identification result.

Claims (6)

1. A high-sulfur-content gas well casing defect identification method based on a direct-current magnetic field and integrated learning is characterized by comprising the following steps:
s1: and scanning and detecting the defective casing of the high-sulfur-content gas well by using a direct-current magnetic field sensor to obtain a static magnetic signal of the casing defect, and marking according to the actual defect type.
S2: the static magnetic signal acquired in step S1 is preprocessed.
S3: and (4) performing data shift on the static magnetic signal preprocessed in the step (S2) by using a moving average method, and realizing the phase alignment of the distorted parts under the same defect type.
S4: the data phase-aligned in step S3 is subjected to feature extraction by principal component analysis.
S5: based on the features extracted in step S5 and their corresponding defect types, a multi-classifier is trained. And combining the classification results of the weak classifiers in a weighted average mode by utilizing a bagging algorithm of ensemble learning.
S6: and (4) processing the high-sulfur-content gas well casing to be tested in the steps S1-S4, extracting corresponding characteristics, inputting the characteristics into the multi-classifier trained in the step S5, and obtaining a model-predicted high-sulfur-content gas well casing defect identification result.
2. The method for identifying the defects of the high-sulfur-content gas well casing based on the direct-current magnetic field and the integrated learning as claimed in claim 1, wherein in the step S1, the defect types comprise: single-sided extrusion, double-sided extrusion, bending, transverse seam, longitudinal seam, hole, four-sided extrusion and no defect.
3. The method for identifying the casing defect of the high-sulfur gas well based on the direct-current magnetic field and the integrated learning as claimed in claim 1, wherein in step S3, static magnetic signal data are circularly shifted, pearson correlation coefficients under different shifting conditions are respectively calculated, and the shift length under the minimum pearson correlation coefficient is taken as the optimal shift length to obtain the static magnetic signal after data shifting.
4. The method for identifying the defects of the high-sulfur gas well casing based on the direct-current magnetic field and the ensemble learning as claimed in claim 1, wherein in the step S2, the preprocessing comprises smoothing filtering, data centralization and the like.
5. The method for identifying the casing defect of the high-sulfur-content gas well based on the direct-current magnetic field and the ensemble learning as claimed in claim 1, wherein in the step S5, the weak classifier is an SVM classifier or a decision tree classifier.
6. The method for identifying the casing defect of the high-sulfur-content gas well based on the direct-current magnetic field and the ensemble learning as claimed in claim 5, wherein in the step S5, the multiple classifiers include 2n weak classifiers, the weak classifiers are binary models, specifically n SVM classifiers and n decision trees, and n is the number of defect types. Each weak classifier predicts the probability of a particular defect type and other defect types, respectively, and each defect type is predicted by an SVM classifier and a decision tree classifier, respectively. And performing weighted summation on the probability of each defect type based on the prediction probability of the weak classifier, and finally performing normalization through a softmax function to output the probability of each defect type, wherein the output with the highest probability is the defect type.
CN202111435387.XA 2021-11-29 2021-11-29 High-sulfur-content gas well casing defect identification method based on direct-current magnetic field and integrated learning Pending CN114154539A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111435387.XA CN114154539A (en) 2021-11-29 2021-11-29 High-sulfur-content gas well casing defect identification method based on direct-current magnetic field and integrated learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111435387.XA CN114154539A (en) 2021-11-29 2021-11-29 High-sulfur-content gas well casing defect identification method based on direct-current magnetic field and integrated learning

Publications (1)

Publication Number Publication Date
CN114154539A true CN114154539A (en) 2022-03-08

Family

ID=80784271

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111435387.XA Pending CN114154539A (en) 2021-11-29 2021-11-29 High-sulfur-content gas well casing defect identification method based on direct-current magnetic field and integrated learning

Country Status (1)

Country Link
CN (1) CN114154539A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20220178242A1 (en) * 2020-12-09 2022-06-09 Baker Hughes Oilfield Operations Llc Identification of wellbore defects using machine learning systems

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20220178242A1 (en) * 2020-12-09 2022-06-09 Baker Hughes Oilfield Operations Llc Identification of wellbore defects using machine learning systems
US11939858B2 (en) * 2020-12-09 2024-03-26 Baker Hughes Oilfield Operations Llc Identification of wellbore defects using machine learning systems

Similar Documents

Publication Publication Date Title
Miao et al. Online defect recognition of narrow overlap weld based on two-stage recognition model combining continuous wavelet transform and convolutional neural network
CN110470729B (en) Eddy current-based nondestructive testing method for oil field casing pipe defects
CN101737039B (en) Multi-functional underground electromagnetic crack detector
CN103997315B (en) A kind of pipeline magnetic flux leakage signal adaptive filter device and method
CN109886433A (en) The method of intelligent recognition city gas pipeline defect
CN114154539A (en) High-sulfur-content gas well casing defect identification method based on direct-current magnetic field and integrated learning
US20220221826A1 (en) System and method for managing wellsite event detection
CN103927528A (en) Coal and rock recognition method based on close neighborhood pixel gray level joint distribution characteristics
CN115359015A (en) Oil well pipe wall thickness logging data interpretation and corrosion defect intelligent determination method
CN114799610A (en) Welding quality real-time detection method and system based on inverse Fourier transform and self-encoder
CN110939428B (en) Identification method for tight sandstone oil and gas reservoir cracks
CN117388359A (en) In-service drill rod thread defect nondestructive testing method based on ultrasonic guided wave
CN113550741A (en) Method for detecting minimum inner diameter of casing
CN202256121U (en) Device for testing and contrasting corrosion of underground oil pipe
CN110792425B (en) Method for measuring water content of formation fluid
CN112282730A (en) Real-time monitoring and evaluating method for deformation of underground casing induced by reservoir fracturing modification
CN113638392A (en) Device for marine static sounding and sampling in same hole
CN112487952A (en) Mine microseismic signal automatic identification method based on deep learning
CN112698424A (en) Method for obtaining fault rock displacement pressure based on rock debris
CN104730122A (en) Underground oil gas detection method based on electronic nose
WO2021034944A1 (en) Systems and methods for core data shifting
Tang et al. Automated monitoring of pilot tube microtunneling installations through pattern recognition in time-series data of hydraulic pressure
US12024993B2 (en) Detecting defects in tubular structures
US20230167732A1 (en) Detecting Defects in Tubular Structures
CN117214958B (en) Advanced geological detection sensing and forecasting system based on long-distance horizontal directional drilling

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination