A defect positioning and identifying method for nondestructive testing of a steel wire rope

Abstract

The application discloses a kind of steel wire rope nondestructive testing defect positioning identification method, mainly including probe structure design, acquisition computing terminal design and defect positioning identification algorithm design.Probe structure design avoids equipment shaking, suitable for different direct steel wire rope detection scene.Acquisition computing terminal design designs the steel wire rope detection terminal equipment for complex field.Defect positioning identification algorithm is based on the collaborative analysis of slice image and slice waveform of local defect LF and metal cross-sectional area loss LMA, according to LMA data verification for LF diagnosis result, combined with respective confidence, to obtain the final steel wire rope defect specific position and type.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of nondestructive testing, and particularly relates to a defect positioning and identification method for nondestructive testing of a steel wire rope. BACKGROUND

[0002] The steel wire rope is widely used in the mining, hoisting, elevator, bridge, cableway, port and ocean industries. Its safety is directly related to the safety of personnel, equipment and hoisted objects. For a long time, due to the complex structure of the steel wire rope, the harsh use environment and the complex stress change in the working process, the steel wire rope often has phenomena such as fatigue, corrosion, wear, broken wire and even rupture, which causes the steel wire rope to often fail to reach the expected service life and be replaced, thereby increasing the use cost. In China, the method of manual visual inspection and periodic forced replacement of the steel wire rope is generally used to ensure safety in production. This method not only discards the steel wire rope that still has use value, causing great waste, but also cannot replace the steel wire rope that is severely damaged due to various accidental factors in time, thereby forming an accident hazard.

[0003] GB / T 21837-2008 Electromagnetic Testing Method for Ferromagnetic Steel Wire Rope divides the defects of the steel wire rope into two categories according to the nature of the defects: Local Flaw (LF) and Loss of Metallic cross-sectional Area (LMA). At present, the LF and LMA signals obtained based on the electromagnetic testing method are still manually checked, which is not only low in efficiency, but also has a huge labor cost. SUMMARY

[0004] To solve the above technical problems, the present application provides a defect positioning and identification method for nondestructive testing of a steel wire rope, which is based on an electromagnetic testing method, uses a target detection algorithm to position the defect position and judge the defect type, and performs labeling and visualization, thereby having good reference significance.

[0005] To achieve the above purpose, the technical solution adopted by the present application is as follows:

[0006] The defect positioning and identification method for nondestructive testing of the steel wire rope comprises a probe structure design, a collection and calculation terminal design and a defect positioning and identification algorithm design. The probe structure design is based on an electromagnetic measurement method and is suitable for different steel wire rope detection scenes. The collection and calculation terminal design realizes data reasoning and calculation. The defect positioning and identification algorithm performs collaborative analysis based on the slice images and slice waveforms of the LF signal and the LMA signal, obtains the final specific position and type of the defect of the steel wire rope according to the diagnosis result of the LF signal, the LMA signal data verification and the respective confidence levels.

[0007] Furthermore, in the probe structure design, four guide wheels 1 are provided to clamp the wire rope according to the different thickness of the wire rope to avoid the shaking of the detection equipment during measurement, which causes fluctuations in the LF signal and LMA signal data; an encoder is installed on one guide wheel to record the actual length of the detected wire rope; the main part of the detection equipment has a permanent magnet and a Hall device inside, which are used to synchronously measure the LF signal and the LMA signal; the probe is provided with an upper device and a lower device, the upper device includes a Hall device, a permanent magnet and an aviation probe, and the lower device includes a Hall device and a permanent magnet, which are connected by a middle buckle; the top of the detection device is provided with an aviation connector to lead out the LF signal and the LMA signal.

[0008] Furthermore, the acquisition and calculation terminal design is provided with an explosion-proof power supply module, a 4G communication module, an aviation interface, a multi-channel differential acquisition unit, a signal noise reduction and filtering circuit module, an Intel CPU motherboard and accessories, and a GPU accelerator card; the explosion-proof power supply module supplies power to the acquisition terminal; the aviation connector is used for data communication between the Hall device and the acquisition card; the multi-channel acquisition module is used to drive the device to perform acquisition tasks; the noise reduction and filtering circuit performs signal processing on the collected data; the Intel CPU motherboard is a signal processing unit; the GPU accelerator card is hung on the Intel CPU motherboard to accelerate the image processing speed.

[0009] Furthermore, the defect location and identification algorithm specifically includes the following steps:

[0010] Step 1: Start the probe, pull the wire rope, and perform Hilbert-Huang transform on the synchronized LF and LMA signals acquired by the probe to achieve signal noise reduction filtering; slice the LF and LMA signals in time series; plot the data points of each slice one by one, draw a two-dimensional curve image of each slice, hide the corresponding horizontal and vertical coordinates, and obtain slice images of the LF and LMA signals, realizing the transformation of the one-dimensional time series signal into a two-dimensional image;

[0011] Step 2: The defect location and recognition algorithm is divided into two parts: model training and model testing. In the model training, the slice images of the LF signal and the LMA signal are used to draw defect boxes and identify the defect types according to the known defect locations and types. Then, the target detection algorithm is used to train the slice images of the LF signal and the slice images of the LMA signal respectively to obtain the target detection models of the LF and LMA signals respectively.

[0012] In the model detection, the probe acquires LF signals and LMA signals each time, slices and reconstructs them into images, and then uses the trained LF target detection model to detect the slice images of the LF signals.

[0013] Further, in the step 1, the slice length is self-defined, and each slice length is basically consistent.

[0014] Further, in the model detection of the step 2, if a defect is identified, according to the defect position given by the target detection model, it is found whether there is a maximum value or a minimum value near the same position of the LMA signal, if there is no maximum or minimum value at the same position of the LMA signal, the defect at the position is determined to be non-existent, if there is a maximum value or a minimum value, the slice image of the LMA signal is identified using the LMA target detection model; when the defect position of the slice image of the LF signal and the defect position of the slice image of the LMA signal are less than 100 mm, there is a suspected defect at the position; the confidence of the position given by the LF target detection model and the LMA target detection model is used for judgment, and the specific judgment is as follows:

[0015]

[0016] Wherein, 0 represents a suspected defect, 1 represents a mild defect, 2 represents a serious defect, P LF Refers to the confidence of the LF image defect position, P LMA Refers to the confidence of the LMA image defect position.

[0017] Advantages of the present application:

[0018] 1. Unlike various optical methods based on visual detection for defect recognition, due to the poor on-site environment, the camera image effect is poor, which seriously affects the detection effect. The present application is based on electromagnetic detection method, uses target detection algorithm to locate the defect position and judge the defect type, and performs labeling visualization, which has good reference significance.

[0019] 2. By fusing LF and LMA signals, the accuracy and reliability of detection are improved. At the same time, manual detection is replaced, and the enterprise operation labor cost is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 It is a probe structure design drawing;

[0021] Figure 2 It is a collection and calculation terminal design drawing;

[0022] Figure 3 It is a defect positioning and identification algorithm flow chart. DETAILED DESCRIPTION

[0023] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0024] This paper designs and implements a method for locating and identifying wire rope defects based on electromagnetic detection and an image recognition algorithm. The method primarily includes the design of a probe structure, an acquisition and calculation terminal, and a defect location and identification algorithm. This method utilizes a target detection algorithm to collaboratively analyze localized damage (LF) and loss of metal area (LMA) signals for defect location and identification.

[0025] The design concept of the present invention is:

[0026] For nondestructive testing of wire rope defects, electromagnetic detection methods are used to obtain the wire rope's LF and LMA signals. However, in actual rope pulling, the signal amplitude varies with factors such as speed. Using threshold detection methods is often prone to false detections and missed detections, significantly reducing the accuracy of the nondestructive testing system and failing to pinpoint the specific fault. Therefore, the present invention fuses the LF and LMA signals, converting the time-series signals into a two-dimensional image for target detection. This achieves high-precision detection of wire rope defects, effectively addressing the low accuracy of wire rope nondestructive testing.

[0027] like Figure 1 As shown, the probe structure design of the present invention is based on the electromagnetic measurement method, and the detection equipment includes a guide wheel 1, a buckle 2, a Hall device 3, a permanent magnet 4, an aviation joint 5, and an encoder 6. The guide wheel 1 is located on both sides of the equipment and is used to tighten the wire rope. The buckle 2 is located in the center of the equipment and is used to tighten the upper and lower parts of the equipment. The Hall device 3 is located inside the equipment and is used to measure the magnetic flux of the wire rope. The permanent magnet 4 surrounds the Hall device and is used to generate a magnetic field. The aviation joint 5 is used to transmit data, and the encoder 6 is located on the guide wheel 1 and is used to record the distance the device moves in the forward or reverse direction. The present invention is designed with four guide wheels 1, which can clamp the wire rope according to the different thickness of the wire rope to avoid the shaking of the detection equipment during measurement, resulting in fluctuations in the LF signal and LMA signal data. At the same time, an encoder 6 is installed on a guide wheel 1 to record the actual length of the wire rope 0 detected by the equipment. The main part of the detection equipment has a permanent magnet 4 and a Hall device 3 inside, which are used to synchronously measure the LF signal and the LMA signal. The upper device includes a Hall effect device 3, a permanent magnet 4, and an aviation probe 5. The lower device also includes a Hall effect device 3 and a permanent magnet 4, which are connected by a middle buckle 2. The top of the detection device is equipped with an aviation connector 5 to lead out the LF signal and the LMA signal.

[0028] like Figure 2 As shown, the acquisition and calculation terminal of the present invention mainly includes an explosion-proof power supply module, a 4G communication module, an aviation interface, a multi-channel differential acquisition unit, a signal noise reduction and filtering circuit module, an Intel CPU motherboard and accessories, and a GPU accelerator card. The explosion-proof power supply module supplies power to the acquisition terminal. Considering the presence of dust in the application acquisition, it needs to be explosion-proof. The aviation connector is used for data communication between the Hall device and the acquisition card. The multi-channel acquisition module is used to drive the device to perform acquisition tasks. Taking into account the on-site noise interference, a noise reduction and filtering circuit is added to perform signal processing on the collected data. The Intel CPU motherboard is the main signal processing unit of the device. The GPU accelerator card is hung on the CPU motherboard to accelerate the image processing speed.

[0029] The CPU calculation of the acquisition and computing terminal mainly relies on the Intel CPU motherboard. In order to accelerate the data inference process, adding a GPU accelerator card helps to quickly realize defect target detection and analysis.

[0030] like Figure 3 As shown, the defect location and identification algorithm of the present invention utilizes multi-dimensional fusion of LF and LMA signals, and locates and identifies defects through a target detection algorithm, which greatly improves the accuracy of wire rope defect identification. The LF and LMA timing signals collected by the probe are easily affected by speed and operating conditions during operation, resulting in changes in their background noise. In particular, during the start-up acceleration and stop-deceleration processes of the wire rope, the curve noise differs significantly, and this part of the noise often exceeds the amplitude of some defects, resulting in unstable defect detection and low accuracy. This has always been one of the difficulties and pain points of non-destructive testing of wire ropes.

[0031] To this end, the present invention designs a set of defect location and recognition algorithms based on the characteristics of LF signals and LMA signals, which specifically includes the following steps:

[0032] Step 1: Start the probe and pull the wire rope. Perform a Hilbert-Huang transform (HHT) on the synchronized LF and LMA signals acquired by the probe to achieve signal noise reduction filtering. Then, slice the LF and LMA signals in a time series manner. The slice length can be customized. For example, the LF signal can be sliced ​​into segments every 1 meter, or cross-overlapping slices can be made. It is sufficient to ensure that the length of each slice is basically the same. The data points of each slice are plotted one by one, and a two-dimensional curve image of each slice is drawn. The corresponding horizontal and vertical coordinates are hidden to obtain the slice image of the LF and LMA signals, realizing the transformation of the one-dimensional time series signal into a two-dimensional image.

[0033] Step 2, the defect location recognition algorithm is divided into two parts: model training and model detection. In the model training, the slice images of the LF signal and the LMA signal are drawn with a defect frame and identified with a defect type according to the known defect position and type. Then the slice images of the LF signal and the LMA signal are trained respectively by using a target detection algorithm to obtain the respective target detection models of the LF and the LMA.

[0034] In the model detection, the probe acquires the LF signal and the LMA signal each time, and after slice reconstruction into images, the trained LF target detection model is used to detect the slice images of the LF signal. If a defect is recognized, according to the defect position given by the target detection model, it is found whether there is a maximum or minimum value near the same position of the LMA signal. If there is no maximum or minimum value at the same position of the LMA signal, the defect at this position is determined to be nonexistent. If there is a maximum or minimum value, the LMA target detection model is used to identify the slice images of the LMA signal. When the defect position of the slice images of the LF signal and the defect position of the slice images of the LMA signal are less than 100 mm, there is a suspected defect at this position. At this time, the confidence of the position given by the LF target detection model and the LMA target detection model is used for judgment, and the specific judgment is as follows:

[0035]

[0036] Wherein, 0 represents a suspected defect, 1 represents a mild defect, 2 represents a serious defect, P LF refers to the confidence of the LF image defect position, P LMA refers to the confidence of the LMA image defect position.

[0037] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for locating and identifying defects in nondestructive testing of wire ropes, characterized by: It includes probe structure design, data acquisition and calculation terminal design, and defect location and identification algorithm design. The probe structure design is based on electromagnetic measurement methods and is suitable for detecting wire ropes of different diameters. The acquisition and calculation terminal is designed to realize data reasoning and calculation; The defect location and identification algorithm is based on the collaborative analysis of the slice images and slice waveforms of the LF signal and the LMA signal. The LF signal diagnosis result is verified according to the LMA signal data and combined with their respective confidence levels to obtain the final specific location and type of the wire rope defect. The defect location and recognition algorithm specifically includes the following steps: Step 1: Start the probe, pull the wire rope, and perform Hilbert-Huang transform on the synchronized LF and LMA signals acquired by the probe to achieve signal noise reduction filtering; slice the LF and LMA signals in time series; plot the data points of each slice one by one, draw a two-dimensional curve image of each slice, hide the corresponding horizontal and vertical coordinates, and obtain slice images of the LF and LMA signals, realizing the transformation of the one-dimensional time series signal into a two-dimensional image; Step 2: The defect location and recognition algorithm is divided into two parts: model training and model testing. In the model training, the slice images of the LF signal and the LMA signal are used to draw defect boxes and identify the defect types according to the known defect locations and types. Then, the target detection algorithm is used to train the slice images of the LF signal and the slice images of the LMA signal respectively to obtain the target detection models of the LF and LMA signals respectively. In the model detection, the probe acquires LF signals and LMA signals each time, slices and reconstructs them into images, and then uses the trained LF target detection model to detect the slice images of LF signals; In the model detection, if a defect is identified, based on the defect location given by the target detection model, it is determined whether there is a maximum or minimum value near the same location of the LMA signal. If there is no maximum or minimum value at the same location of the LMA signal, the defect at that location is deemed not to exist. If there is a maximum or minimum value, the slice image of the LMA signal is identified using the LMA target detection model. When the defect location of the slice image of the LF signal and the defect location of the slice image of the LMA signal are less than 100 mm, there is a suspected defect at that location. The confidence level of the location given by the LF target detection model and the LMA target detection model is used to make a judgment, and the specific judgment is as follows: Among them, 0 represents suspected defects, 1 represents mild defects, 2 represents severe defects, and P LF Refers to the confidence of the defect location in the LF image, P LMA Refers to the confidence of the defect location in the LMA image.

2. A method for locating and identifying defects in nondestructive testing of steel wire ropes according to claim 1, characterized in that: In the probe structure design, four guide wheels are set to clamp the wire rope according to the different thickness of the wire rope to avoid shaking of the detection equipment during measurement, which may cause fluctuations in the LF signal and LMA signal data; an encoder is installed on one guide wheel to record the actual length of the detected wire rope; a permanent magnet and a Hall device are provided inside the main body of the detection equipment to synchronously measure the LF signal and the LMA signal; the probe is provided with an upper device and a lower device, the upper device includes a Hall device, a permanent magnet and an aviation probe, and the lower device includes a Hall device and a permanent magnet, which are connected by a middle buckle; the top of the detection device is provided with an aviation connector to lead out the LF signal and the LMA signal.

3. The method for locating and identifying defects in nondestructive testing of steel wire ropes according to claim 1, characterized in that: The acquisition and computing terminal design includes an explosion-proof power supply module, a 4G communication module, an aviation interface, a multi-channel differential acquisition unit, a signal noise reduction and filtering circuit module, an Intel CPU motherboard and accessories, and a GPU accelerator card. The explosion-proof power supply module provides power to the acquisition terminal. The aviation interface is used for data communication between the Hall effect device and the acquisition card. The multi-channel differential acquisition unit is used to drive the device to perform acquisition tasks; the signal noise reduction and filtering circuit module performs signal processing on the collected data; the Intel CPU motherboard is a signal processing unit; the GPU accelerator card is hung on the Intel CPU motherboard and is used to accelerate the image processing speed.

4. The method for locating and identifying defects in nondestructive testing of steel wire ropes according to claim 1, characterized in that: In step 1, the slice length is customized, and the length of each slice is basically the same.

Images