Method for rapidly upgrading AI function of traditional eddy current detection instrument, detection system and detection instrument

Abstract

The invention discloses a method and system for rapidly upgrading an AI function of a traditional eddy current detection instrument, and the method comprises the steps: collecting an image of an eddy current detection impedance planar graph track displayed on a display screen of an existing eddy current detection instrument in real time through a camera which is fixedly arranged above the display screen of the eddy current detection instrument; and in combination with an AI discrimination model, identifying the morphological characteristics of the eddy current detection impedance planar graph track, discriminating whether the morphological characteristics of the track of the image are matched with the morphological characteristics of the track of a preset defect signal, and accordingly discriminating whether the current eddy current detection signal is a defect signal, thereby realizing accurate classification of defects. The intelligent AI function of a traditional eddy current detection instrument is transformed and upgraded quickly at low cost, and the detection precision is improved.

Description

Technical Field

[0001] The present invention relates to the field of nondestructive testing technology, and in particular to a method for quickly adding an AI function to a traditional eddy current detector, a detection system, and a detection instrument. Background Art

[0002] Eddy current nondestructive testing, a core testing method in modern industry, is widely used in the metallurgical industry for real-time monitoring of online tubes, rods, and wires. However, the testing equipment currently used in the metallurgical industry is all traditional eddy current testing equipment. Alarm mechanisms generally employ amplitude / phase alarm thresholds, or even rely on amplitude alarm functions. Their defect determination logic is based on fixed thresholds and cannot dynamically distinguish between true defects and interference signals. For example, in the automated inspection of metallurgical tubes, through-holes in tubes and rods are defects that require detection. However, inherent material noise, indentations, and deformation significantly impact the eddy current testing impedance plane signal. The threshold range of these detection signals is similar to that of through-holes, and the trajectory morphology on the impedance plane is similar. Therefore, these false defect signals are easily misinterpreted as through-hole defect signals by existing eddy current instruments, affecting the accuracy of the test results. In industrial practice, the upgrade and transformation of such equipment faces bottlenecks. Traditional intelligent upgrades require hardware or software modifications to existing equipment, which is not only costly and time-consuming, but also seriously impacts production line continuity.

[0003] Based on this, how to quickly, cheaply and non-invasively realize the transformation and upgrading of intelligent AI functions without changing the core hardware architecture of traditional eddy current testing instruments has become an urgent problem to be solved in the current industry. Summary of the Invention

[0004] To solve the above problems, the present invention provides a method for rapidly upgrading the AI ​​function of a traditional eddy current testing instrument. The present invention is implemented as follows:

[0005] A method for rapidly upgrading traditional eddy current testing instruments to AI capabilities, including:

[0006] The image of the eddy current detection impedance plane diagram trajectory displayed on the display screen of the eddy current detection instrument is collected in real time by a camera fixedly arranged above the display screen of the eddy current detection instrument;

[0007] Inputting the collected image into an AI discrimination model, wherein the AI ​​discrimination model is trained to identify the morphological features of the eddy current detection impedance plane map trajectory;

[0008] Analyzing the input image using the AI ​​discrimination model to identify whether the trajectory morphology features of the image match the trajectory morphology features of the preset defect signal, and based on this, determining whether the current eddy current detection signal is a defect signal;

[0009] If it is determined to be a defect signal, an alarm is triggered; if it is determined to be a non-defect signal, no alarm is triggered or a prompt different from the defect alarm is triggered.

[0010] Furthermore, the present invention is applied to the intelligent discrimination of through-hole defect signals and surface indentation signals in the online automatic eddy current testing of pipelines, wherein:

[0011] The AI ​​discrimination model is trained to identify the sequential features of trajectory direction changes in the image;

[0012] The preset defect signal is a through-hole defect signal of a pipeline, and its corresponding trajectory morphological feature is preset as a direction change sequence of "first descending and then ascending";

[0013] The AI ​​discrimination model also presets the surface indentation signal of the pipeline as a non-defect signal, and presets its corresponding trajectory morphological feature as a direction change sequence of "first rising and then falling"; after identifying the "first rising and then falling" trajectory morphological feature, the AI ​​discrimination model determines it as a non-defect signal.

[0014] Furthermore, the AI ​​discrimination model is a convolutional neural network model;

[0015] The training data of the convolutional neural network model includes:

[0016] A sample set of trajectory images labeled as “through-hole defects” with a “first descending then ascending” direction change sequence characteristic;

[0017] A set of trajectory image samples labeled as "surface indentation" with the characteristic of "rising first and then falling" direction change sequence;

[0018] A sample set of random trajectory images labeled “proper noise”.

[0019] The present invention also discloses an eddy current detection system, comprising an eddy current detection instrument, a probe and an alarm device, and further comprising:

[0020] An image acquisition unit, comprising a camera fixedly mounted above the display screen of the eddy current testing instrument, for acquiring in real time an image of the trajectory of the eddy current testing impedance plane diagram displayed on the display screen of the eddy current testing instrument;

[0021] an AI discrimination unit, communicatively connected to the image acquisition unit and the alarm control module of the alarm device, wherein the AI ​​discrimination unit is provided with the trained AI discrimination model as described above;

[0022] The AI ​​identification unit is configured to:

[0023] receiving an image of the eddy current detection impedance plane map trajectory from the image acquisition unit;

[0024] Identifying trajectory morphological features of the image using the AI ​​discrimination model;

[0025] Based on the identified trajectory morphological features, determine whether the current eddy current detection signal is a defect signal;

[0026] A control signal is sent to the alarm control module according to the discrimination result.

[0027] Furthermore, the camera is installed through an adjustable bracket to ensure that the optical axis of the camera is perpendicular to the plane of the eddy current detection instrument display screen, and the field of view of the camera completely covers the impedance plane diagram display area on the eddy current detection instrument display screen.

[0028] Furthermore, the AI ​​identification unit is integrated on the main control circuit board of the eddy current detection instrument; or, the AI ​​identification unit is deployed in an external computing device or cloud server that is communicatively connected to the eddy current detection instrument.

[0029] The present invention also discloses an eddy current detection instrument: comprising the eddy current detection system as described in any one of the above items, and the camera is integrated with the display screen housing of the eddy current detection instrument.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The solution of the present invention uses an external camera to capture images of the eddy current detection impedance plane trajectory on the display screen of a traditional eddy current testing instrument. Combined with an AI discrimination model, it distinguishes between the defect signal to be tested and the noise defect signal. There is no need to modify the core hardware components of the traditional eddy current testing instrument, that is, there is no need to eliminate existing equipment, which extends the equipment life. There is no need for complex system integration and debugging, shortening the upgrade cycle. Overall, it avoids the high cost and lengthy modification cycle of traditional intelligent upgrades, and significantly reduces investment and upgrade costs.

[0032] The present invention adopts an AI discrimination model to objectively and accurately identify the morphological characteristics of the impedance plane diagram trajectory of eddy current detection, effectively distinguish defect signals from non-defect signals, reduce human misjudgment, and improve the accuracy of detection results; it is particularly suitable for the intelligent discrimination of through-hole defect signals and surface indentation signals in online automatic eddy current detection of pipelines in the metallurgical industry. By identifying the sequential characteristics of trajectory direction changes, it can accurately distinguish between these two similar but essentially different signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the examples of the present invention or the technical solutions in the prior art or the drawings required for the description of the prior art, a brief introduction is given. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0034] Figure 1 These are two pipeline defect signal trajectory diagrams.

[0035] Figure 2 Schematic diagram of the detection process of the present invention. DETAILED DESCRIPTION

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention.

[0037] like Figure 1 As shown, the impedance graph trace of signal A first descends and then ascends, indicating a typical through-hole defect in a pipe. The impedance graph trace of signal B, on the other hand, first ascends and then descends, indicating a signal generated by an indentation on the pipe surface. In actual service, this does not affect normal operation and is generally not treated as a defect. However, conventional automated pipe and rod defect detection systems in the metallurgical industry employ amplitude / phase alarm thresholds. Consequently, non-defect features such as indentations and deformations can be easily misidentified as through-holes or cracks due to phase similarity. Furthermore, minor defects, such as early-stage cracks, are often overlooked because their signal amplitudes are below the threshold, potentially developing into defects that affect the normal operation of the pipe and rod.

[0038] In this embodiment, to address the problems of insufficient intelligence and high misjudgment rate in current eddy current detectors, a method for rapidly upgrading traditional eddy current detection instruments to AI functions is provided. Without modifying the existing eddy current instrument hardware and software, by adding the now mature AI image recognition function, multimodal data fusion and dynamic defect classification detection are achieved, thereby improving detection accuracy and efficiency. The specific method includes:

[0039] By using a camera fixedly arranged above the display screen of the eddy current detection instrument, images of the eddy current detection impedance plane diagram trajectory displayed on the display screen of the eddy current detection instrument are collected in real time; by using an external camera to collect display screen images, there is no need to modify the internal circuit or signal processing module of the eddy current detector, thus avoiding the high cost and technical threshold of traditional hardware upgrades, and is particularly suitable for the intelligent transformation of old equipment; and the camera is fixed above the display screen and does not touch the internal signal link of the instrument. This non-invasive compatible design effectively avoids interference with the original detection function and ensures the stability and reliability of the detection.

[0040] In this embodiment, an industrial camera with a resolution of 1920×1080 at 60fps and an exposure time of 1 / 100s is used to avoid screen refresh streaks that may occur with existing instruments. White balance is manually adjusted to prevent color distortion caused by automatic adjustment. An anti-shake bracket is used to position the industrial camera 30cm above the display screen of the existing eddy current detector, ensuring a perpendicularity error of ≤1° between the lens optical axis and the display plane. Furthermore, a polarizing filter is added in front of the camera lens to eliminate display screen reflections, obtaining clear images and reducing the workload of image post-processing.

[0041] The collected image is input into an AI discrimination model, which is trained to identify the morphological features of the eddy current detection impedance plane diagram trajectory; the AI ​​model is used to analyze the impedance plane diagram trajectory morphology in real time, replacing manual observation and experience-based reading and judgment, thereby improving the efficiency and accuracy of defect identification and reducing misjudgments and missed judgments.

[0042] Analyzing the input image using the AI ​​discrimination model to identify whether the trajectory morphology features of the image match the trajectory morphology features of the preset defect signal, and based on this, determining whether the current eddy current detection signal is a defect signal;

[0043] If a signal is identified as a defect, an alarm is triggered. If a signal is not identified as a defect, no alarm is triggered or a different prompt is triggered. Triggering differentiated alarms based on the identification results helps operators respond quickly and optimize the inspection process.

[0044] Furthermore, the present invention is applied to the intelligent discrimination of through-hole defect signals and surface indentation signals in online automatic eddy current testing of pipelines. It mainly targets two easily confused signals, through-hole defects and surface indentation, which are common in online pipeline testing. The intelligent discrimination is achieved by changing the trajectory direction sequence, solving the technical problem that traditional methods are difficult to effectively distinguish.

[0045] The AI ​​discrimination model is trained to identify the sequential features of trajectory direction changes in the image;

[0046] The preset defect signal is a through-hole defect signal of a pipeline, and its corresponding trajectory morphological feature is preset as a direction change sequence of "first descending and then ascending";

[0047] The AI ​​discrimination model also pre-determines that the surface indentation signal of the pipeline is a non-defect signal and that its corresponding trajectory morphological characteristics follow a directional sequence of "rising first, then falling." Upon identifying this "rising first, then falling" trajectory morphological characteristic, the AI ​​discrimination model determines it as a non-defect signal. The direct correlation between trajectory morphological characteristics and the physical properties of the defect type imbues the AI ​​discrimination model with clear engineering logic and enhances its interpretability.

[0048] Furthermore, the AI ​​discrimination model is a convolutional neural network model;

[0049] The training data of the convolutional neural network model includes:

[0050] A sample set of trajectory images labeled as “through-hole defects” with a “first descending then ascending” direction change sequence characteristic;

[0051] A set of trajectory image samples labeled as "surface indentation" with the characteristic of "rising first and then falling" direction change sequence;

[0052] A sample set of random trajectory images labeled “proper noise”.

[0053] The core architecture of the convolutional neural network model specifically includes a multi-level feature extraction module and a classification decision module, among which:

[0054] The multi-level feature extraction module consists of two convolutional layers. The first uses a 5x5 convolution kernel to extract global morphological features of the impedance map captured by the video head, including trajectory direction and curvature, with a particular focus on trajectory direction. The second convolutional layer uses a 3x3 convolution kernel to capture local details, such as turning point mutations. Each layer is followed by a ReLU activation function and a 2x2 max pooling function to form a hierarchical feature abstraction.

[0055] In the classification decision module, the fully connected layer flattens the feature map into a vector with an input dimension of 64*54*54 and an output dimension of 256. The output layer uses a 3-neuron structure, corresponding to the class probabilities of through-hole, indentation, and noise, respectively.

[0056] The present invention also discloses an eddy current detection system, comprising an eddy current detection instrument, a probe and an alarm device, and further comprising:

[0057] An image acquisition unit, comprising a camera fixedly mounted above the display screen of the eddy current testing instrument, for acquiring in real time an image of the trajectory of the eddy current testing impedance plane diagram displayed on the display screen of the eddy current testing instrument;

[0058] An AI discrimination unit, which is in communication with the image acquisition unit and the alarm control module of the alarm device, can be adapted to different models of eddy current testing instruments, and contains the trained AI discrimination model as described above;

[0059] The AI ​​identification unit is configured to:

[0060] receiving an image of the eddy current detection impedance plane map trajectory from the image acquisition unit;

[0061] Identifying trajectory morphological features of the image using the AI ​​discrimination model;

[0062] Based on the identified trajectory morphological features, determine whether the current eddy current detection signal is a defect signal;

[0063] A control signal is sent to the alarm control module according to the discrimination result.

[0064] In this embodiment, the alarm control module includes multi-level alarms. The first-level alarm corresponds to through-hole defects, and is set to trigger the red LED to flash and superimpose a buzzer sound; the second-level alarm corresponds to non-through-hole defects such as indentations, and is set to trigger the yellow LED to be always on without a sound alarm. The background can generate a detection log containing a timestamp and defect image.

[0065] Furthermore, the camera is installed through an adjustable bracket to ensure that the optical axis of the camera is perpendicular to the plane of the eddy current detection instrument display screen, support on-site rapid calibration, adapt to display screens of different sizes, and reduce installation complexity; and the field of view of the camera completely covers the impedance plane diagram display area on the eddy current detection instrument display screen, ensuring that the trajectory image is not missed, thereby improving the accuracy of subsequent AI analysis.

[0066] Furthermore, the AI ​​discrimination unit is integrated on the main control circuit board of the eddy current detection instrument; or, the AI ​​discrimination unit is deployed in an external computing device or cloud server that is communicatively connected to the eddy current detection instrument. The deployment method can be selected according to the requirements of the detection scenario: local integration is suitable for offline or low-latency scenarios, and cloud deployment is suitable for big data analysis and deep learning model iteration.

[0067] The present invention also discloses an eddy current testing instrument comprising the eddy current testing system described in any of the above items, wherein the camera is integrated with the display housing of the eddy current testing instrument. This integrated design eliminates the need for external device cables and space requirements, improving the instrument's portability and operational ease. It also enhances the instrument's vibration and interference resistance, making it suitable for complex industrial environments such as pipeline inspection. Furthermore, the integrated design facilitates mass production and market promotion, promoting the upgrade of traditional eddy current testing instruments to intelligent and automated ones.

[0068] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for rapidly upgrading the AI ​​function of a traditional eddy current testing instrument, characterized in that: include: The image of the eddy current detection impedance plane diagram trajectory displayed on the display screen of the eddy current detection instrument is collected in real time by a camera fixedly arranged above the display screen of the eddy current detection instrument; Inputting the collected image into an AI discrimination model, wherein the AI ​​discrimination model is trained to identify the morphological features of the eddy current detection impedance plane map trajectory; Analyzing the input image using the AI ​​discrimination model to identify whether the trajectory morphology features of the image match the trajectory morphology features of the preset defect signal, and based on this, determining whether the current eddy current detection signal is a defect signal; If it is determined to be a defect signal, an alarm is triggered; if it is determined to be a non-defect signal, no alarm is triggered or a prompt different from the defect alarm is triggered.

2. The method for rapidly upgrading the AI ​​function of a conventional eddy current testing instrument according to claim 1, characterized in that: Applied to the intelligent discrimination of through-hole defect signals and surface indentation signals in the online automatic eddy current testing of pipelines, including: The AI ​​discrimination model is trained to identify the sequential features of trajectory direction changes in the image; The preset defect signal is a through-hole defect signal of a pipeline, and its corresponding trajectory morphological feature is preset as a direction change sequence of "first descending and then ascending"; The AI ​​discrimination model also presets the surface indentation signal of the pipeline as a non-defect signal, and presets its corresponding trajectory morphological feature as a directional change sequence of "first rising and then falling"; after identifying the "first rising and then falling" trajectory morphological feature, the AI ​​discrimination model determines it as a non-defect signal.

3. The method for rapidly upgrading the AI ​​function of a conventional eddy current testing instrument according to claim 2, characterized in that: The AI ​​discrimination model is a convolutional neural network model; The training data of the convolutional neural network model includes: A sample set of trajectory images labeled as "through-hole defects" with a "first descending then ascending" direction change sequence characteristic; A set of trajectory image samples labeled as "surface indentation" with the characteristic of "rising first and then falling" direction change sequence; A sample set of random trajectory images labeled "property noise".

4. An eddy current detection system, comprising an eddy current detection instrument, a probe and an alarm device, adopting the detection method according to claim 1, characterized in that: Also includes: An image acquisition unit, comprising a camera fixedly mounted above the display screen of the eddy current testing instrument, for acquiring in real time an image of the trajectory of the eddy current testing impedance plane diagram displayed on the display screen of the eddy current testing instrument; An AI discrimination unit, communicatively connected to the image acquisition unit and the alarm control module of the alarm device, wherein the AI ​​discrimination unit is provided with a trained AI discrimination model; The AI ​​identification unit is configured to: receiving an image of the eddy current detection impedance plane map trajectory from the image acquisition unit; Identifying trajectory morphological features of the image using the AI ​​discrimination model; Based on the identified trajectory morphological features, determine whether the current eddy current detection signal is a defect signal; A control signal is sent to the alarm control module according to the discrimination result.

5. The eddy current detection system according to claim 4, characterized in that: The camera is installed through an adjustable bracket to ensure that the optical axis of the camera is perpendicular to the plane of the eddy current detection instrument display screen, and the field of view of the camera completely covers the impedance plane diagram display area on the eddy current detection instrument display screen.

6. The eddy current detection system according to claim 4, characterized in that: The AI ​​identification unit is integrated on the main control circuit board of the eddy current testing instrument; or, The AI ​​identification unit is deployed in an external computing device or a cloud server that is communicatively connected to the eddy current detection instrument.

7. An eddy current testing instrument, characterized in that: The eddy current detection system comprises the eddy current detection system according to any one of claims 4 to 6, wherein the camera is integrated with the display screen housing of the eddy current detection instrument.

Images