A hidden defect shape and position quantitative method and system based on multi-modal thermal feature fusion

Abstract

The application discloses a hidden defect shape and position quantitative method and system based on multi-modal thermal feature fusion. The method comprises the following steps: establishing an electromagnetic-thermal-defect multi-physical field coupling model, and calculating an analytical relationship between the defect hiding depth and the feature disengagement time under the echo path and the feature peak time under the straight-through path; adopting an echo and straight-through path cooperative detection scheme, and synchronously collecting the surface transient temperature field of the test piece by using double infrared thermal imagers; pre-processing and normalizing the temperature field data, and extracting multi-dimensional thermal features in the time domain and the space domain; based on the analytical relationship, calculating the defect hiding depth by using the time domain features, determining the transverse and longitudinal positions of the defect by using the space domain features, and comprehensively evaluating the equivalent geometric size of the defect by fusing information; finally, performing three-dimensional visualization on the results and generating a detection report. By means of the multi-modal thermal feature fusion, the quantitative evaluation of the hiding depth, position and size of the hidden defect is realized, and the "panoramic" information of the defect is provided.

Description

Technical Field

[0001] This invention belongs to the field of advanced nondestructive testing and intelligent evaluation technology, specifically involving a method and system that integrates electromagnetic induction thermal imaging, multiphysics modeling and feature machine learning to achieve high-precision quantitative "seeing through" the depth, location and size of hidden defects inside materials. Background Technology

[0002] In the manufacturing and maintenance of high-end equipment such as aerospace, nuclear power, and rail transportation, accurate quantitative assessment of hidden defects within components is crucial to ensuring structural integrity and service safety. Electromagnetic induction thermal imaging technology is widely used due to its advantages such as non-contact and full-field detection.

[0003] However, the mainstream application of this technology is still at the qualitative or semi-quantitative level, and the following technical bottlenecks exist: 1. Insufficient quantification capability: Most existing methods rely on the experience of inspectors to interpret thermal images, making it difficult to accurately extract geometric parameters such as the hidden depth, absolute position, and true size of defects, thus failing to meet the "measurement-level" industrial requirements.

[0004] 2. Insufficient understanding of the model mechanism: The electromagnetic-thermal-defect multi-physics coupling effect is complex, and there is a lack of a general model that can accurately describe the analytical relationship between thermal response characteristics and defect shape and position parameters, resulting in a weak quantitative theoretical foundation.

[0005] 3. Limited detection mode: It relies on a single infrared thermal imaging method, which limits its detection capabilities, makes it insensitive to deep defects, has insufficient defect resolution, and lacks an effective mechanism for multi-mode information complementarity and fusion.

[0006] 4. Poor anti-interference ability: The thermal signal is easily affected by uneven heating, material property fluctuations, surface conditions and environmental noise, resulting in poor robustness and low reliability of the quantification results of existing methods.

[0007] Therefore, developing an innovative method and technology system that can achieve high-precision, automated, and visualized quantitative evaluation of the geometric parameters of hidden defects has become an urgent technical need in this field. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method and system for quantitative and visual evaluation of hidden defects through multimodal thermal feature fusion. This invention aims to achieve a technological leap from "defect detection" to "precise defect quantification" by constructing a precise mapping model between "time-space-spectrum" multidimensional thermal features and defect shape and position parameters.

[0009] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method and system for quantitative analysis of hidden defect shape and position through multimodal thermal feature fusion. The method is characterized by achieving precise quantification of defects by constructing a quantitative mapping model between defect shape and position parameters and "time-space-spectrum" multidimensional thermal features, comprising the following steps: S1. Definition of Multiphysics Analytical Modeling and Quantization Criteria A physical model of electromagnetic-thermal-defect multiphysics coupling is established, and the defect hiding depth is derived based on this model. L r ) and characteristic separation time under echo path ( t r ) and characteristic peak time under the direct path ( t t The universal analytical relationship between them forms the core quantitative criteria: Echo path: ,in Represents the thermal conductivity of the material. For density, Specific heat capacity.

[0010] Straight path: ,in Represents the thermal conductivity of the material. For density, Specific heat capacity.

[0011] S2, Multimodal Collaborative Data Acquisition A collaborative detection scheme combining echo and direct path is adopted: the electromagnetic induction excitation unit applies thermal excitation to the specimen; a dual infrared thermal imager system triggered by a synchronous controller is used to synchronously acquire the transient temperature fields of the excitation side surface under the echo path and the back surface under the direct path, respectively, to obtain a spatiotemporally aligned complementary thermal response sequence.

[0012] S3, Intelligent Extraction and Fusion Inversion of Multidimensional Thermal Features S31. Preprocessing and Normalization: The acquired raw thermal image sequence is time-aligned, noise filtered, and then... Normalization is performed to eliminate the influence of excitation fluctuations and surface emissivity.

[0013] S32. Multidimensional Feature Extraction: Automatically extract the following feature sets from normalized hot data: Temporal characteristics: including characteristic disconnection time under the echo path. t r Characteristic peak time under direct path t t .

[0014] Spatial characteristics include the centroid coordinates of the thermal anomaly region (for lateral positioning), equivalent diameter, area change rate, and heat flow propagation morphology obtained through a three-dimensional thermal flow field reconstruction algorithm.

[0015] Frequency domain features: By performing a fast Fourier transform on the temperature time series signal, the phase or amplitude features of a specific frequency band can be extracted.

[0016] S33, Geometric Parameter Fusion Inversion: Deep quantization: extracting t r and t t Substituting the analytical relationships in S1 respectively, the estimated burial depth under echo and straight path conditions is calculated, and the final accurate burial depth is obtained through weighted fusion. L r .

[0017] Location calibration: The lateral position of the defect is determined using the centroid coordinates in the spatial domain features; combined with depth... L r The longitudinal location of the defect in the thickness direction is determined by combining the heat flow diffusion pattern with the overall determination of the defect.

[0018] Size assessment: Based on the equivalent diameter and area in the spatial domain features, and introducing the correlation model between the "defect width-to-depth ratio" and the "thermal response grayscale ratio", a width-to-depth ratio of less than 2 is set as a reliable detection threshold, thereby assessing the equivalent width and height of the defect.

[0019] S4, 3D Visualization and Digital Report Generation The defect shape and position parameters obtained by inversion are integrated and spatially located and visualized in the three-dimensional computer-aided design model of the specimen. A digital inspection report containing the three-dimensional defect model, accurate shape and position parameters and confidence level is generated. This report can be directly integrated into the industrial digital twin platform.

[0020] Secondly, the present invention provides a system for implementing the above method, characterized in that the system comprises: 1. Collaborative Incentive and Data Acquisition Module: Electromagnetic induction excitation unit: includes a high-frequency induction heating power supply and an excitation coil.

[0021] Dual thermal imager synchronous acquisition unit: includes at least two infrared thermal imagers and a synchronization controller, used to realize synchronous data acquisition of echo and straight path.

[0022] 2. Core Module for Quantitative Analysis: Computing and Control Unit: A high-performance computer workstation, which includes: Multiphysics analytical model library: stores quantization criteria as described in S1.

[0023] Multidimensional hot feature automatic extraction algorithm package: used to perform the feature extraction described in S32.

[0024] Defect shape and position parameter fusion and inversion program: used to perform parameter calculation and fusion as described in S33.

[0025] Self-learning optimization unit: It can continuously optimize the inversion model using historical detection data through deep learning networks to improve quantization accuracy.

[0026] 3. Intelligent output and visualization module: Results output unit: Used to generate and display standardized digital reports containing a 3D visualization model of defects, as described in S4.

[0027] Compared with the prior art, the present invention has the following significant advantages: 1. A leap from "qualitative" to "precise quantitative": By establishing the law-correlation between "characteristic time and hidden depth", the physical model-driven direct calculation of the hidden depth of defects is realized, rather than empirical estimation, which significantly improves the accuracy.

[0028] 2. A multi-dimensional information fusion evaluation system was constructed: the temporal, spatial and spectral multi-dimensional thermal features were fused to achieve comprehensive and high-precision quantification of defect depth and location, providing a "panoramic" information of defects.

[0029] 3. Improved robustness and universality of the system: Through normalization processing and data-driven feature selection and machine learning optimization, various interference factors are effectively suppressed, enabling the method to maintain reliable quantization performance under different materials and working conditions.

[0030] 4. Achieved visualization and digital integration of test results: The results are directly mapped to the 3D model and a digital report is generated, which greatly improves the intuitiveness of the results and the integration capability with modern intelligent manufacturing systems, providing an accurate data foundation for predictive maintenance and life assessment. Attached Figure Description

[0031] Figure 1 Multimodal thermal characteristic defect evaluation system; Figure 2 Transient temperature response of different defects under two different paths; Figure 3 Thermal imaging sequence of defects 1-4 along the straight path; Figure 4 Thermal image sequence of defects 1-4 under echo path; Figure 5 Normalized transient temperature response of defects at different longitudinal positions under two different paths; Figure 6Thermal image sequence of defects 5-7 under the echo path; Figure 7 Thermal image sequence of defects 5-7 along the straight path; Figure 8 Normalized transient temperature response of defects at different lateral positions along the echo path; Figure 9 Thermal image sequences of defects at three different locations under the echo path; Figure 10 Thermal image sequences of defects in three locations along a straight path. Detailed Implementation

[0032] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0033] When an excitation signal is applied to the excitation coil near the conductor material under test, eddy currents are induced on the conductor surface. The Joule heat generated by these eddy currents is conducted into the material's interior and lower surface, forming a heat flow field. When this heat flow encounters defects on the interior or lower surface, its conduction process is disrupted, thus affecting the temperature distribution on the surface of the material under test. A multimodal thermal characteristic defect evaluation system, such as a thermal imager and temperature measuring instrument, records and analyzes the surface temperature changes of the material under test. Figure 1 As shown. The power supply parameters of the electromagnetic induction excitation device are: voltage 60V, current 50A, excitation frequency 200kHz, and output power 3kW. The specimen is kept parallel to the planar coil, and the heating time is 40ms. The dual thermal imagers acquire data at a frame rate of 50Hz, and timing synchronization is ensured by a synchronous controller.

[0034] 1. Evaluation of Defect Hiding Depth Normalized transient temperature response of defects at different hiding depths, such as Figure 2 As shown, the temperature curve of the defect-free region is a decreasing curve after heating. When a defect exists on the back side of the specimen, heat flow is disturbed when heat is conducted to the defect area. The heat is reflected, causing the temperature above the defect to rise, and its temperature curve will deviate from the temperature curve of the defect-free region. The greater the defect's hidden depth, the later the temperature curve deviates from the normal temperature curve, and the longer its deviance time. Therefore, the deviance time is used to quantify the hidden depth of the defect. Figure 2 The response curve for a defect width of 4mm shows that the smaller the defect hiding depth, the shorter the time for heat to reach the defect, and the earlier the temperature deteriorates.

[0035] By observing the transient temperature response along the direct path, it was found that the smaller the defect's hidden depth, the earlier the peak appeared. Therefore, this peak time can be used to quantify the defect's hidden depth.

[0036] A rectangular specimen with a thickness of 10 mm has a cylindrical defect with a diameter of 5 mm. The heights of defects 1, 2, 3, and 4 are 3 mm, 5 mm, 7 mm, and 9 mm, respectively. The thermal imaging sequence of defects 1, 2, 3, and 4 along a straight path is as follows: Figure 3 As shown. Figure 3 In the thermal image shown in (a) at 40ms, i.e., the end of the heating phase, only defect 4 shows a significant temperature rise, and the resulting high-temperature region is clearly visible. Figure 3 (b) In the 200ms thermal image of the cooling phase, defect 2 shows a sequential increase in temperature. Figure 3 (c) shows a 400ms thermal image. As heat conduction proceeds, the high-temperature region formed by defect 2 is larger than... Figure 3 (b) More obvious. Figure 3 As shown in the 600ms thermal image (d), defects 1-4 are all present. This demonstrates a direct correlation between the depth of defect concealment and the timing of the appearance of the high-temperature region. The later the high-temperature region appears, the greater the depth of defect concealment.

[0037] Defect thermal images under echo path as shown Figure 4 As shown. Figure 4 (a) is the thermal image at 600ms, where only the high-temperature area of ​​defect 4 is clearly visible. Figure 4 (b) is a thermal image at 800ms, showing high-temperature areas in all four defects. Figure 4 As can be seen in (c) and (d), the high-temperature regions of the four defects gradually stabilize and become more pronounced. This indicates that the later the high-temperature region appears, the greater the hidden depth of the defect. Therefore, by analyzing the time of the first appearance of the high-temperature region in the thermal imaging sequence, the hidden depth of the defect can be effectively quantified.

[0038] 2. Evaluation of the longitudinal location of hidden defects To determine the longitudinal position of the defect, the transverse position of the defect is fixed at the transverse center, and the normalized transient temperature response of the defect at different distances from the bottom surface is observed.

[0039] Under the echo path, the normalized transient temperature response curves of defects at different locations are as follows: Figure 5 As shown in the left figure, the greater the distance between the defect and the bottom surface of the specimen, the earlier the temperature shift occurs and the faster the shift speed. The figure shows that defects farther from the bottom surface than 2.5 mm are easier to detect and distinguish, and their longitudinal location is more accurate. For shallow defects, the separation time method can be used preferentially. For deep defects, it is necessary to combine direct path data or other data for analysis.

[0040] Under a straight path, the normalized transient temperature response of defects at different longitudinal positions is as follows: Figure 5As shown in the right figure, observing the entire temperature response process reveals that, after normalization, the transient temperature response curves of defects at different locations along the straight path exhibit roughly the same trend. The peak times of all curves in the figure show small differences, but subtle variations exist between them. A magnified view of the 500ms-1000ms range in the figure reveals more significant differences in the temperature responses of defects at different locations in this section. Defects with a distance greater than 2.5mm from the bottom surface are easily distinguishable, and their locations are readily identifiable. When the distance from the bottom surface is less than 2.5mm, the curves in the figure overlap considerably and are difficult to differentiate. By introducing a specific time point (600ms), it is found that the greater the distance from the bottom surface (i.e., the shallower the hiding depth), the higher the temperature.

[0041] For the detection of the longitudinal location of internal defects in the specimen, defects 5, 6, and 7 are cylindrical in shape with a diameter of 5 mm, a height of 3 mm, and hidden depths of 6.5 mm, 3.5 mm, and 0.5 mm, respectively. The thermal image sequences of defects 5, 6, and 7 under the echo path are as follows: Figure 6 As shown. Figure 6 (a)-(d) are thermal images at 600ms, 800ms, 1000ms and 1200ms, respectively. Figure 6 (a) Defect 7, with the smallest hidden depth, first appears in a high-temperature region within 600ms. As time progresses, heat conduction occurs... Figure 6 (b) The high-temperature region of defect 6 is more pronounced in 800ms compared to 15(a), while the high-temperature region of defect 5 remains indistinct. Figure 6 The process sequence shows a correlation between the defect's hidden depth and the timing of the high-temperature region's appearance; the earlier the high-temperature region appears, the shallower the defect's hidden depth. This characteristic can be used to infer the defect's longitudinal orientation.

[0042] Thermal image sequences of defects 5, 6, and 7 under the direct path are as follows: Figure 7 As shown. By Figure 7 It can be seen that no obvious high-temperature areas appeared at any of the three longitudinal defects at different times, and there was no phenomenon of high-temperature areas gradually appearing over time. For the straight path, the back side is farther from the induction coil, and it takes a longer time for the induced heat to be conducted to the back side. Careful observation... Figure 7 Even so, internal defects can still be detected. Furthermore, defects at different longitudinal locations show different temperatures on the back side. For example... Figure 7 (c) From left to right, as the distance between the defect and the back surface increases, the temperature image of the defect gradually becomes blurred, which is also the basis for identifying the longitudinal position of the defect.

[0043] 3. Evaluation of the lateral location of hidden defects To determine the lateral position of the defect, the defect is fixed at the longitudinal center, and the normalized transient temperature response of the defect at different distances from the side is observed.

[0044] Under the echo path, the transient temperature response of defects at different transverse locations is as follows: Figure 8 As shown in the left figure. Defects (1) and (9), (2) and (8), (3) and (7), and (4) and (6) are all symmetrical, while defect (5) is located at the center of the specimen. Figure 8 As shown in the left figure, the transient temperature response curves of the symmetrically distributed defects almost overlap. The closer to the center, the slower the temperature decreases at the data monitoring point.

[0045] In the straight path, the transient temperature response of defects at different lateral positions is as follows: Figure 8 As shown in the right figure, the pattern shown in the figure is consistent with the echo path. The closer to the center, the easier it is to detect, and the data collected for symmetrical defects are basically consistent.

[0046] However, both echo paths and straight paths can only distinguish whether a defect is near the center, but cannot distinguish whether the defect is on the left or right. Of course, the lateral orientation of the defect can be distinguished by offsetting the temperature monitoring point.

[0047] For the detection of the lateral location of internal defects in the specimen, the defects were cylinders with a diameter of 5 mm, a height of 3 mm, and a hiding depth of 3.5 mm. For comparison of the lateral location of the defects, a 24 mm * 24 mm square area was used, with the defects located at the left, center, and right positions of the observation area. The thermal image sequences of the defects at the three locations under the echo path are as follows: Figure 9 As shown. Figure 9 The four time points are 300ms, 400ms, 600ms, and 1000ms. Comparing the thermal images of defects at the three locations at these four times, the high-temperature area of ​​the defect is not obvious before 400ms, a significant high-temperature area appears at 600ms, and the high-temperature area is more obvious at 1000ms than at 600ms. Under the echo path, the volume of the high-temperature region caused by the internal defect is larger than the actual volume of the defect. For example... Figure 9 (a) The defect on the left side is clearly visible in the thermal images at four different times. The defect area on the left side is significantly hotter than the non-defect area, making it easier to identify the location of the defect. Figure 9 The same pattern is observed in the thermal image of the defect on the right side of (c). Therefore, within a small observation area, if a defect exists around the area, it will also be reflected in the thermal image. The high-temperature area will gradually become more apparent over time; by selecting an appropriate time, the lateral location of the defect can be determined.

[0048] Under a straight path, thermal imaging sequences of defects at three locations are as follows: Figure 10 As shown. Figure 10 The four time points are 1200ms, 1400ms, 1600ms, and 1800ms. During the cooling period, the temperature response caused by the defect becomes increasingly apparent over time. Within a relatively small observation area, defects close to the observation area can also be observed.

[0049] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method and system for quantitative analysis of hidden defect shape and position based on multimodal thermal feature fusion, characterized in that, The method includes the following steps: S1. Multiphysics Modeling and Quantization Criterion Definition: Establish a physical model of electromagnetic-thermal-defect multiphysics coupling, and derive the defect hiding depth based on this model. L r ) and characteristic separation time under echo path ( t r ) and characteristic peak time under the direct path ( t t The analytical relationship between them; S2. Multimodal collaborative data acquisition: A collaborative detection scheme combining echo and direct path is adopted, and a synchronously triggered dual infrared thermal imager system is used to synchronously acquire the transient temperature field of the specimen surface under the two paths. S3. Intelligent extraction and fusion inversion of multidimensional thermal features: S31. Preprocessing and normalization: Processing the acquired transient temperature field data; S32. Multidimensional feature extraction: Extracting multidimensional thermal features, including temporal and spatial features; S33. Form and position parameter fusion and inversion: Based on the analytical relationship, the defect hiding depth is calculated using time domain features; the lateral and longitudinal positions of the defect are determined using spatial domain features; and the equivalent geometric size of the defect is comprehensively evaluated by fusing information from the echo and the straight path. S4. 3D Visualization and Digital Report Generation: Integrates quantitative evaluation results, spatially locates and visualizes defects in the 3D model of the specimen, and generates a digital inspection report.

2. The method according to claim 1, characterized in that, The analytical relation is specifically as follows: Defect hiding depth under the echo path L r Time of decoupling from features t r satisfy: ,in Represents the thermal conductivity of the material. For density, Specific heat capacity; Defect hiding depth under a direct path L r With characteristic peak time t t satisfy: ,in Represents the thermal conductivity of the material. For density, Specific heat capacity.

3. The method according to claim 1, characterized in that, The intelligent extraction and fusion inversion step of the multidimensional thermal features also includes normalization processing of the original thermal image sequence, using a normalization factor of 1. ,in Differential temperature, L For the specimen thickness, Joule heating is generated on the surface.

4. The method according to claim 1, characterized in that, The extraction of spatial features includes analyzing the heat flow propagation pattern using a three-dimensional thermal flow field reconstruction algorithm to determine the centroid coordinates (for lateral positioning), equivalent diameter, and area change rate of the thermal anomaly region.

5. The method according to claim 1, characterized in that, The comprehensive evaluation of the equivalent geometric dimensions of the defect includes establishing a correlation model between the "defect width-to-depth ratio" and the "thermal response grayscale ratio", and setting a width-to-depth ratio of less than 2 as a reliable detection threshold.

6. The method according to claim 1, characterized in that, In the intelligent extraction and fusion inversion step of multidimensional thermal features, a feature dimensionality reduction and selection algorithm based on principal component analysis is introduced to automatically select the feature combination most sensitive to the shape and position parameters.

7. A system for implementing the method according to any one of claims 1 to 6, characterized in that, The system includes: The collaborative excitation and acquisition module includes an electromagnetic induction excitation unit and a dual infrared thermal imager system triggered by a synchronous controller; The core module for quantitative analysis is the calculation and control unit, which includes a multi-physics analytical model library, an automatic extraction algorithm for multi-dimensional thermal features, and a defect shape and position parameter fusion and inversion program. The intelligent output and visualization module includes a result output unit for generating digital reports containing a 3D visualization model of the defects.

8. The system according to claim 7, characterized in that, The quantitative analysis core module is further equipped with a self-learning optimization unit, which can continuously optimize the inversion process using historical detection data through a machine learning model.

Images