Steel bar corrosion detection method and system based on passive infrared guidance and ultrasonic quantification

Abstract

The application discloses a steel bar corrosion detection method based on passive infrared guidance and ultrasonic quantification, and relates to the technical field of civil engineering structure health monitoring; the method comprises the following steps: acquiring an original infrared thermal image sequence of a target concrete structure surface collected by a passive infrared thermal imager; identifying a thermal anomaly area caused by steel bar corrosion in the infrared thermal image sequence, and determining the physical space coordinates of the thermal anomaly area on the target concrete structure surface; acquiring an ultrasonic time domain signal of the target concrete structure surface collected by an ultrasonic detection assembly at the physical space coordinates; and inputting the ultrasonic time domain signal into a pre-trained corrosion rate prediction model to obtain a steel bar corrosion rate detection result of the target concrete structure surface. The application can realize hierarchical detection of infrared guidance preliminary screening and ultrasonic fixed-point quantification, and can balance the wide-range general survey efficiency and the quantitative evaluation accuracy of local damage.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring technology in civil engineering, and in particular to a method and system for detecting steel corrosion based on passive infrared guidance and ultrasonic quantification. Background Technology

[0002] Reinforced concrete structures, as the core framework of modern infrastructure, directly impact public safety and long-term economic benefits. Steel corrosion is a primary factor leading to structural performance deterioration, making early screening and precise quantitative assessment a crucial requirement in the engineering field.

[0003] Currently, commonly used non-destructive testing (NDT) techniques in engineering practice suffer from two significant limitations. Firstly, electrochemical methods, such as the half-cell potential method, while sensitive to early corrosion reactions, typically require micro-drilling of the concrete cover to connect reinforcing bars. This invasive process necessitates point-by-point contact, making it unsuitable for rapid surveys of large-scale facilities like long-span bridges and tall structures. Secondly, physical acoustic methods, such as ultrasonic methods, while accurately quantifying internal physical damage, suffer from low detection efficiency, require dense grid scanning across the entire area, heavily rely on manual feature extraction (such as wave velocity and amplitude), and exhibit poor generalization ability in complex noise environments.

[0004] Furthermore, while infrared thermal imaging technology, as a non-contact, wide-field-of-view detection method, boasts extremely high survey efficiency, its application is currently largely limited to qualitative screening, making it susceptible to misjudgments due to interference from non-corrosion defects (such as construction honeycombing and voids). Therefore, improving the quantitative assessment accuracy of steel reinforcement corrosion damage while ensuring non-contact, low-cost surveys has become an urgent technical challenge. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a method and system for detecting steel corrosion based on passive infrared guidance and ultrasonic quantification. It can achieve graded detection by infrared-guided initial screening and ultrasonic point-to-point quantification, taking into account both the efficiency of large-scale general surveys and the accuracy of quantitative assessment of local damage.

[0006] To achieve the above objectives, the present invention provides a method for detecting steel corrosion based on passive infrared guidance and ultrasonic quantification, comprising: Obtain the original infrared thermal image sequence of the target concrete structure surface acquired by a passive infrared thermal imager; Identify the thermal anomaly region caused by steel corrosion in the infrared thermal image sequence, and determine the physical spatial coordinates of the thermal anomaly region on the surface of the target concrete structure. Acquire the ultrasonic time-domain signal of the target concrete structure surface at the physical space coordinates by the ultrasonic detection component; The ultrasonic time-domain signal is input into a pre-trained corrosion rate prediction model to obtain the steel corrosion rate detection result of the target concrete structure surface.

[0007] Optionally, identifying thermal anomaly regions in the infrared thermal image sequence caused by steel corrosion includes: Candidate regions exhibiting abnormal temperature gradients in the infrared thermal imaging sequence were selected. Candidate areas with geometric features that are strip-shaped and whose distribution direction is consistent with the predicted direction of the steel reinforcement arrangement are identified as thermal anomaly areas caused by steel reinforcement corrosion.

[0008] Optionally, determining the physical spatial coordinates of the thermal anomaly region on the surface of the target concrete structure includes: Determine the pixel coordinates of the thermal anomaly region in the corresponding image of the infrared thermographic sequence; Based on preset spatial resolution calibration parameters, the pixel coordinates are converted into physical spatial coordinates of the target concrete structure surface.

[0009] Optionally, the corrosion rate prediction model includes: The input layer is used to receive the ultrasonic time-domain signal; The feature extraction module is used to extract the local echo features of the ultrasonic time-domain signal; An attention mechanism module is used to adaptively weight the local echo features to enhance the signal features related to the reflection from the steel reinforcement interface and suppress the background noise features caused by the scattering of concrete aggregate. The regression prediction module is used to map adaptively weighted local echo features to steel corrosion rate detection results.

[0010] Optionally, the method further includes: The steel reinforcement corrosion rate detection results are associated and stored with the corresponding physical space coordinates.

[0011] Optionally, the method further includes: Based on the steel corrosion rate detection results and physical space coordinates, a corrosion rate distribution map is generated in real time on the visualization interface.

[0012] Optionally, the method further includes: Based on the preset corrosion rate grading threshold, the corrosion damage level of each corrosion area in the steel bar corrosion rate detection results is determined, and a corresponding level of early warning is triggered.

[0013] Optionally, the ultrasonic time-domain signal is input into a pre-trained corrosion rate prediction model, including: The ultrasonic time-domain signal is preprocessed; the preprocessing includes bandpass filtering, amplitude normalization, and data length standardization. The preprocessed ultrasonic time-domain signal is input into the pre-trained corrosion rate prediction model.

[0014] This invention also provides a steel reinforcement corrosion detection system based on passive infrared guidance and ultrasonic quantification, comprising: The infrared image acquisition unit is used to acquire the original infrared thermal image sequence of the target concrete structure surface acquired by the passive infrared thermal imager. A thermal anomaly identification and localization unit is used to identify thermal anomaly areas caused by steel corrosion in the infrared thermal image sequence and determine the physical spatial coordinates of the thermal anomaly areas on the surface of the target concrete structure. An ultrasonic signal acquisition unit is used to acquire ultrasonic time-domain signals of the target concrete structure surface collected by the ultrasonic detection component at the physical space coordinates. The corrosion rate prediction unit is used to input the ultrasonic time-domain signal into a pre-trained corrosion rate prediction model to obtain the detection result of the steel corrosion rate on the surface of the target concrete structure.

[0015] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: The steel corrosion detection method based on passive infrared guidance and ultrasonic quantification provided by this invention uses a passive infrared thermal imager to perform non-contact full-field scanning of the structural surface. Without the need for any drilling or point-by-point contact operations on the concrete protective layer, it can quickly acquire thermal image sequences of a large area of ​​the surface under the excitation of a natural heat source. Thus, it achieves efficient survey of large-scale infrastructure while ensuring the non-invasiveness and low cost of the detection process.

[0016] Based on this, by identifying thermal anomaly areas caused by steel corrosion in infrared thermographic sequences, temperature anomaly features specifically associated with steel corrosion damage can be extracted in a targeted manner. This effectively reduces misjudgments caused by non-corrosion defects such as construction honeycomb and voids, and delineates key areas of concern for subsequent accurate quantitative detection. At the same time, the physical spatial coordinates corresponding to the thermal anomaly areas are determined, directly transforming the results of large-scale screening into precise spatial location guidance, significantly reducing the blindness of ultrasonic testing and the workload of scanning.

[0017] After obtaining spatial coordinate guidance, ultrasonic detection components are used to acquire signals at the corresponding locations. The acquired ultrasonic time-domain signals carry specific damage information regarding the reflection and scattering at the rebar interface, providing highly sensitive acoustic data for quantitative analysis. By inputting the acquired ultrasonic time-domain signals into a pre-trained corrosion rate prediction model, the model's ability to automatically extract deep waveform features directly outputs the rebar corrosion rate detection results for each measurement point, achieving an objective and accurate quantitative assessment of the degree of rebar corrosion. This invention's graded detection method, combining infrared-guided positioning and ultrasonic point-to-point quantification, effectively integrates the advantages of both non-destructive testing technologies, balancing the dual needs of large-area rapid screening and quantitative assessment of local damage. Attached Figure Description

[0018] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same parts.

[0019] Figure 1 This is a schematic diagram of the process flow of a steel reinforcement corrosion detection method based on passive infrared guidance and ultrasonic quantification, as shown in an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the passive infrared detection principle of an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the daily temperature difference trend on the surface of a concrete structure, as shown in an embodiment of the present invention. Figure 4 This is a schematic diagram of the interactive positioning principle interface for thermal anomaly regions, as shown in an embodiment of the present invention. Figure 5 This is a schematic diagram of the ultrasonic detection component arrangement for the single-sided planar measurement method according to an embodiment of the present invention; Figure 6 This is a schematic diagram showing the comparison of ultrasonic echo waveforms of steel bars in intact and corroded states, as illustrated in an embodiment of the present invention. Figure 7 This is a schematic diagram of the network architecture of the corrosion rate prediction model shown in an embodiment of the present invention; Figure 8 This is a schematic diagram of the module structure of a steel corrosion detection system based on passive infrared guidance and ultrasonic quantification, as shown in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 , Figure 1 This is a schematic diagram of the process for detecting steel corrosion based on passive infrared guidance and ultrasonic quantification.

[0022] A method for detecting steel corrosion based on passive infrared guidance and ultrasonic quantification includes: S101: Acquire the original infrared thermal image sequence of the target concrete structure surface acquired by a passive infrared thermal imager.

[0023] In applications, passive infrared thermal imagers can be used to perform non-contact full-field scanning of the target concrete structure surface. Solar radiation or ambient diurnal temperature variations serve as natural excitation sources for unsteady-state heat conduction processes; the thermal resistance effect is based on the delamination, voids, or micro-crack interfaces within the concrete protective layer caused by the volume expansion of steel corrosion products; and the infrared thermal imager captures the abnormal temperature gradient formed between damaged and intact areas on the concrete structure surface under conditions of impeded heat flow, thereby generating the original infrared thermal image sequence.

[0024] See Figure 2 In reinforced concrete structures, the volume expansion of steel corrosion products can cause microcracks, hollow areas, or delamination within the concrete cover. These damaged areas are filled with air or loose oxides, whose thermal conductivity is much lower than that of the intact concrete matrix, thus forming a "thermal resistance layer" within the structure.

[0025] The difference in thermal resistance can be derived from the one-dimensional Fourier law of thermal conduction: In the formula: k Thermal conductivity; Temperature gradient; q This represents the heat flux density along the direction of heat flow. λ Indicates the thermal conductivity of the medium; ΔT This indicates the temperature difference across the medium. d Indicates the thickness of the dielectric layer; R This indicates the thermal resistance of the dielectric layer.

[0026] This significant difference in thermal resistance hinders the transfer of solar radiation heat inwards, causing localized heat buildup on the surface of the damaged area during the heating phase. Furthermore, because the thermal conductivity of air is much lower than that of the concrete matrix, the thermal resistance in the damaged area is significantly reduced. R Significantly increased: During the heating phase, heat accumulation on the surface of the damaged area causes its temperature to be higher than the surrounding area, forming a hot spot. During the cooling phase, heat dissipation from the damaged area is hindered and it cannot receive deep heat replenishment, causing its temperature to be lower than the surrounding area, forming a cold spot. Infrared thermal imagers receive the infrared energy radiated from the object's surface and convert it into a temperature distribution image, thereby revealing these hidden damages. The temperature measurement principle of infrared thermal imagers is based on the Stefan-Boltzmann law. The total radiative exitance of the object's surface... M Its absolute temperature T It is directly proportional to the fourth power, as shown in the following formula: In the formula: M The emission amplitude; ε The emissivity of the concrete surface is typically 0.85-0.95. σ It is the Stephen-Boltzmann constant; T It is the thermodynamic temperature.

[0027] By measuring radiation energy M By presetting the emissivity, the temperature field distribution on the structure surface can be calculated. T。

[0028] To obtain infrared thermal imaging data with a high signal-to-noise ratio that is conducive to damage identification, the timing of data acquisition is crucial. (See also...) Figure 3 Infrared scanning is best performed during the rapid heating phase after ample sunlight or the rapid cooling phase after sunset, when the thermal contrast between damaged and intact areas is most significant, effectively highlighting the location of hidden damage caused by steel corrosion. Simultaneously, the zero-crossing point between morning and evening should be avoided; during this period, the concrete surface is in a state of equilibrium between heat absorption and release, and the surface temperature of damaged and intact areas tends to be the same, making it difficult for the infrared thermal imager to effectively distinguish damaged areas and easily leading to missed detections. Through the above passive infrared thermal imaging data acquisition method, a large-area, non-contact, full-field scan of large concrete structures can be completed in a relatively short time.

[0029] S102: Identify the thermal anomaly region caused by steel corrosion in the infrared thermal image sequence, and determine the physical spatial coordinates of the thermal anomaly region on the surface of the target concrete structure.

[0030] The identification of thermal anomaly regions caused by steel corrosion in infrared thermal imaging sequences specifically includes: Candidate regions exhibiting abnormal temperature gradients in infrared thermal imaging sequences were selected. Candidate areas with geometric features that are strip-shaped and whose distribution direction is consistent with the predicted direction of the steel reinforcement arrangement are identified as thermal anomaly areas caused by steel reinforcement corrosion.

[0031] In applications, the original infrared thermographic sequence is displayed on a display device. The infrared data can be presented in the form of grayscale images or pseudo-color images for easy observation and analysis. Because the volume expansion of steel corrosion products usually occurs along the length of the steel bar, it causes damage such as longitudinal cracks or delamination and hollowing distributed along the direction of the steel bar inside the concrete cover. These damaged areas often appear in infrared thermographic images with different colors or brightness than the intact areas, showing as areas of abnormal temperature gradients.

[0032] To accurately distinguish between thermal anomaly areas caused by steel reinforcement corrosion and those caused by non-corrosion defects such as construction honeycombing and voids, this invention utilizes the specific morphological characteristics of corrosion damage in its spatial distribution. Thermal anomaly areas caused by corrosion generally exhibit a strip-like distribution, with their orientation highly consistent with the internal steel reinforcement arrangement, forming a linear or mesh-like thermal anomaly profile corresponding to the steel reinforcement mesh. In contrast, non-corrosion defects often appear as isolated points or irregular clumps, lacking a directional distribution pattern. Based on this morphological difference, candidate areas exhibiting abnormal temperature gradients can be screened from infrared thermographic sequences. Furthermore, candidate areas with strip-like geometric features and a distribution orientation consistent with the predicted steel reinforcement arrangement direction are identified as thermal anomaly areas caused by steel reinforcement corrosion, thereby effectively eliminating interference from non-corrosion defects and achieving preliminary localization of suspected corrosion areas.

[0033] Determining the physical spatial coordinates of the thermal anomaly region on the surface of the target concrete structure specifically includes: Determine the pixel coordinates of the thermal anomaly region in the corresponding image of the infrared thermographic sequence; Based on preset spatial resolution calibration parameters, pixel coordinates are converted into physical spatial coordinates of the target concrete structure surface.

[0034] After identifying the thermal anomaly region, its physical spatial coordinates on the target concrete structure surface need to be obtained to provide spatial guidance information for subsequent ultrasonic detection. First, the pixel coordinates of the thermal anomaly region in the infrared thermographic sequence are determined. These pixel coordinates can be determined through manual specification or automatic system identification. In the manual specification method, the geometric centroid pixel coordinates of the region can be determined based on the user's marking of the thermal anomaly region. In the automatic identification method, the thermal anomaly identification and positioning unit automatically extracts the geometric center of the thermal anomaly region as its pixel coordinates based on image processing algorithms. Subsequently, according to pre-calibrated spatial resolution parameters, the pixel coordinates are converted into physical spatial coordinates on the target concrete structure surface. The spatial resolution coefficient can be determined by the infrared thermal imager's shooting distance or a reference object of known actual size in the image; its value represents the actual physical length corresponding to a single pixel in the image. Through the conversion between pixel coordinates and pre-calibrated parameters, the physical spatial coordinates of the thermal anomaly region on the concrete structure surface are finally output, providing a positional basis for the precise placement of the ultrasonic detection components.

[0035] See Figure 4 Taking an intuitive human-computer interaction method as an example, the inspection personnel identify the damaged area based on their experience, and the system assists in recording the location. The specific operation is as follows: The image display transmits the acquired raw infrared thermal image sequence to a display terminal (such as a laptop or tablet). The display terminal directly presents the infrared data as a grayscale image or pseudo-color image for easy human observation.

[0036] Inspectors visually observe the thermal images on the screen. Specifically, to distinguish damage caused by steel reinforcement corrosion from other randomly distributed concrete defects (such as honeycombing and voids), this invention utilizes the morphological characteristics of corrosion damage. Because the volume expansion of steel reinforcement corrosion products typically occurs along the length of the steel reinforcement, it leads to longitudinal cracks or strip-like delamination in the protective layer. Therefore, in infrared thermal images, thermal anomaly areas caused by corrosion often exhibit a linear or mesh-like distribution highly consistent with the internal steel reinforcement mesh. By identifying this specific morphological feature, interference from non-corrosion defects appearing as isolated points or irregular clumps can be effectively eliminated, achieving specific screening of suspected corrosion areas. Due to the thermal resistance effect of corrosion, the damaged area will exhibit a different color or brightness than the surrounding background (i.e., a thermal anomaly area). Inspectors use a mouse or touchscreen to select or click on the identified thermal anomaly area. Upon receiving the inspector's marking instruction, the system automatically locks onto the area.

[0037] The system determines the location of the area in the image based on the markers left by the inspector. If it is a click operation: the system directly reads the pixel coordinates of the click point. u , vThe center point of the rectangle is used as the geometric centroid of the region. If a selection operation is performed: the system automatically calculates the center point of the rectangle's length and width, and uses it as the pixel coordinates of the geometric centroid of the thermal anomaly region. u , v ).

[0038] To guide subsequent ultrasonic testing, pixel coordinates are converted to physical coordinates of the structural surface using a preset scale. Preset scale: Before testing, a simple conversion factor is determined based on the infrared thermal imager's shooting distance or a reference object of known size in the image. K For example, if 100 pixels in the image represent an actual length of 500 millimeters, then... K =5. Coordinate transformation: using the formula X = uK and Y = vK It automatically converts the pixels on the image into their actual distances relative to the origin of the structure. X , Y Output: The screen displays the physical location information of the point for the testing personnel to refer to when placing the ultrasonic probe. The calculation formula is as follows: ( X , Y () represents the physical spatial coordinates of the target point on the surface of the concrete structure; u , v () represents the pixel coordinates of the target point in the infrared thermal image. K This represents the spatial resolution coefficient, or scale, which indicates the actual physical length corresponding to a single pixel in the image. This coefficient is derived from the formula... K = L / P The calculation shows that, among which L The actual length of the reference object. P The pixel length of the reference object in the image; X 0 , Y 0 ), ( u 0 , v 0 ) represents the translation of the origin of the physical coordinate system, used to calibrate the starting positions of the two coordinate systems.

[0039] S103: Acquire the ultrasonic time-domain signal of the target concrete structure surface at physical space coordinates by the ultrasonic detection component.

[0040] Based on the physical space coordinates obtained in the preceding steps for guiding ultrasonic detection, refined ultrasonic data acquisition can be carried out on the same side surface of the target concrete structure to obtain ultrasonic time-domain signals carrying information on steel reinforcement interface damage. To adapt to the actual engineering conditions where steel reinforcement is encased in concrete, this invention abandons the traditional steel reinforcement end transmission method and adopts a single-sided planar measurement method for non-destructive data acquisition.

[0041] The ultrasonic detection component employs a packaged piezoelectric ultrasonic detection system, primarily composed of a transmitting probe and a receiving probe. Probe selection: Piezoelectric ceramic transducers with a center frequency of 50kHz-200kHz are preferred (such as the PZT-4 transmitting type and the PZT-5 receiving type). Ultrasonic waves in this frequency band exhibit both good penetration depth in concrete and scattering sensitivity at steel reinforcement interfaces (diameter 10mm-32mm). Encapsulation structure: The probe is encased in a robust metal or engineering plastic shielding shell, with an acoustic matching layer on the bottom surface to protect the piezoelectric crystal and adapt to rough concrete surfaces.

[0042] See Figure 5 In the detection process involving probe placement and coupling, the inspection personnel or a wall-climbing robot carrying the equipment reach the designated coordinates. X , Y After that, perform the following operations: Single-sided arrangement: The transmitting probe (T) and the receiving probe (R) are placed side by side on the same side surface of the concrete structure; geometric alignment: The probe positions are adjusted so that the line connecting the centers of the two probes crosses the geometric centroid of the suspected corroded area.

[0043] To achieve independent and accurate measurement of individual target steel bars (whether longitudinal or transverse) within a complex steel mesh, this invention designs the following layout strategy: Vertical crossing and locking (target identification): Based on the direction of the rebar identified by infrared image recognition, the transmitting and receiving probes are symmetrically arranged on both sides of the axis of the target rebar, so that the direction of the line connecting the centers of the two probes is perpendicular to the axis of the target rebar, i.e., orthogonal crossing. This arrangement utilizes the strong reflection characteristics of the ultrasonic beam on the cylindrical surface of the target rebar to establish the main signal source.

[0044] Adjacent Interference Avoidance (Eliminating the Influence of Co-directional Reinforcing Bars): Precise centering is achieved using the extracted thermal anomaly centroid coordinates, ensuring the midpoint of the probe connection line is precisely above the target reinforcing bar. The beam spread angle is controlled by combining the optimized probe chip size, concentrating ultrasonic energy to cover the target reinforcing bar. Spatial attenuation characteristics are utilized to weaken the sound beam's impact on adjacent co-directional reinforcing bars. Opposite Interference Avoidance (Eliminating the Influence of Crossing Reinforcing Bars): For opposite reinforcing bars that intersect the target reinforcing bar (e.g., longitudinal bars when measuring transverse direction), avoidance is implemented at the nodes. Based on the infrared grid image, the probe placement is selected in the "grid gap area" between two opposite reinforcing bars, avoiding placing the probe directly above the reinforcing bar intersection node, thus physically cutting off the obstruction and reflection of the sound beam by the opposite reinforcing bars. Fixed Spacing: Maintaining the center distance between the transmitting and receiving probes. L It is a fixed value. A fixed spacing is a prerequisite for accurately locating the reflected wave through subsequent acoustic timing.

[0045] Acoustic coupling: To eliminate the air gap between the probe and the concrete surface, apply an acoustic coupling medium (such as medical petroleum jelly, butter or special ultrasonic coupling paste) to the bottom surface of the probe and apply moderate pressure to ensure that the probe is in close contact with the concrete and to establish a stable sound propagation channel.

[0046] Pulse Excitation: A high-voltage pulse signal is excited by the transmitting probe controlled by an ultrasonic pulse generator. Ultrasonic waves propagate into the concrete as body waves. Waveform Acquisition: The ultrasonic waves penetrate the concrete protective layer and undergo physical interaction at the steel-concrete interface. If the steel is intact, the large difference in acoustic impedance at the interface results in strong specular reflection; if the steel is corroded, the loose rust layer causes a rough interface, resulting in diffuse reflection and scattering, and increased acoustic energy attenuation. Time-Domain Data Output: The receiving probe captures the weak vibration signal reflected back from the interface, which is then converted from analog to digital by a data acquisition card to generate an ultrasonic time-domain waveform sequence. This sequence records the change in voltage amplitude over time, serving as input data for subsequent deep learning models. This is to achieve accurate quantization.

[0047] In practical applications, the preferred technical parameters of the ultrasonic detection component are shown in Table 1.

[0048] Table 1

[0049] Furthermore, the collected data can be pre-analyzed based on the following acoustic theoretical model: Formula for calculating acoustic time: In a single-sided planar layout, ultrasonic waves mainly propagate along a V-shaped path within the concrete. Assume the concrete cover thickness is... H The center-to-center distance between the transmitting probe and the receiving probe is L The speed of ultrasonic waves in concrete is V c .

[0050] According to the principles of geometric acoustics, the theoretical acoustic time for an ultrasonic wave to travel from the transmitting end to the receiving end after being reflected by the steel reinforcement interface is... t TOF It can be calculated using the following formula: In the formula: t TOF This represents the theoretical sound time of an ultrasonic wave returning to the receiver after being emitted and reflected. H Indicates the concrete protective thickness of the reinforcing steel being tested; L This indicates the fixed distance between the center of the transmitting probe and the center of the receiving probe; V c This represents the speed of sound propagating in the concrete medium, and this value can be measured through calibration tests.

[0051] Acoustic impedance and reflection coefficient formula: The reflection intensity of ultrasonic waves reaching the interface of steel reinforcement depends on the difference in acoustic impedance between the two media. Acoustic impedance Z Defined as dielectric density ρ With the speed of sound V The product (i.e.) Z = ρ × V） .

[0052] Sound pressure reflection coefficient at the concrete-reinforcement interface R Defined as: In the formula: R denoted by the sound pressure reflection coefficient, its range is [-1, 1], and the smaller the absolute value, the weaker the reflection; Z steel Indicates the acoustic impedance of the reinforcing steel material; Z conc This indicates the acoustic impedance of the concrete medium.

[0053] Scattering attenuation formula: Besides changes in reflection characteristics, corrosion also leads to increased interface roughness, causing ultrasonic wave scattering attenuation. The amplitude of the signal received at the receiver... A Follows the law of exponential decay: In the formula: A 0 represents the initial transmission amplitude; α ( f () represents the frequency-dependent material attenuation coefficient; x The propagation path length; S ( λ , δ The scattering factor is the interface scattering factor, which depends on the ultrasound. λ Micro-roughness of the interface with corrosion δ As corrosion deepens, roughness... δ Increased scattering effect leads to a decrease in the main peak energy in the time-domain waveform and an increase in clutter components in the tail wave.

[0054] Considering the complexity of concrete as a heterogeneous multiphase material, the actual ultrasonic time-domain signal acquired by the receiving probe... x ( t This can be modeled as the superposition of target echo and structural noise. Its mathematical expression is: In the formula: x ( t The received raw ultrasonic time-domain waveform; s(t) For effective echo components, by K It is composed of superimposed reflected waves from different paths; A k For the first k The amplitude of each echo component is affected by the interface reflection coefficient. R Influence; w(t) Let be the envelope function of the transmitted pulse; n scat (t) This is structural scattering noise caused by coarse aggregate and microcracks.

[0055] See Figure 6 The receiver probe acquires the actual ultrasonic time-domain signal, where the effective target echo component is composed of multiple reflected waves from different paths, and its amplitude is affected by the interface reflection coefficient; while structural noise is mainly caused by non-uniform media such as coarse aggregate and microcracks. The increased roughness of the steel reinforcement interface caused by corrosion will change the reflection and scattering characteristics, resulting in specific changes in the time-domain waveform, such as a decrease in the main peak energy and an increase in the wake clutter component.

[0056] S104: Input the ultrasonic time-domain signal into the pre-trained corrosion rate prediction model to obtain the steel corrosion rate detection result of the target concrete structure surface.

[0057] After acquiring the ultrasonic time-domain signal, it is input into a pre-trained corrosion rate prediction model to obtain the detection result of the steel corrosion rate on the surface of the target concrete structure. This corrosion rate prediction model is a deep learning model that can extract deep features related to the steel corrosion state from the input ultrasonic time-domain signal and establish a precise mapping relationship from ultrasonic waveform to corrosion rate. The model is trained on a large amount of ultrasonic sample data with known corrosion rate labels, possessing the ability to identify the specificity of the reflection waveform at the steel interface, effectively eliminating false positive interference caused by non-corrosion defects, and outputting a quantitative prediction value of the steel corrosion rate at the measurement point location. Through an end-to-end learning approach, the model can automatically focus on key time-domain features in the signal related to steel interface reflection, suppressing background interference caused by environmental noise and concrete aggregate scattering, thereby achieving an accurate quantitative assessment of the degree of steel corrosion.

[0058] When inputting ultrasonic time-domain signals into a pre-trained corrosion rate prediction model, the specific steps include: The ultrasonic time-domain signal is preprocessed; the preprocessing includes bandpass filtering, amplitude normalization, and data length standardization. The preprocessed ultrasonic time-domain signal is input into the pre-trained corrosion rate prediction model.

[0059] In applications, to eliminate environmental noise interference and adapt to the input requirements of subsequent deep learning models, the following preprocessing operations can be performed on the acquired raw ultrasound time-domain signals: Bandpass filtering: Due to the presence of power frequency interference (50Hz) and high-frequency electromagnetic noise in the detection environment, a digital bandpass filter can be designed to address these issues. For example, the lower cutoff frequency can be set to 0.5 times the probe's center frequency, and the upper cutoff frequency to 2.0 times the center frequency, in order to retain the main echo energy and filter out out-of-band noise.

[0060] Amplitude normalization: Considering the different coupling conditions and concrete attenuation levels at different measurement points, the voltage amplitude of the original ultrasonic time-domain signal may vary by orders of magnitude (e.g., some peak values ​​are 5V, while others are 0.5V). To accelerate the convergence of the neural network, amplitude normalization can be performed on each sampling sequence. x ( t Perform Z-score standardization: In the formula: The input sequence is standardized. This is the mean of the signal segment; This represents the standard deviation of the signal segment. The processed signal has zero mean and unit variance, eliminating the influence of the absolute signal strength and allowing the model to focus on changes in waveform morphology (i.e., scattering and reflection characteristics) rather than the magnitude of the voltage.

[0061] Data length normalization: To meet the requirements of the input layer of the model for a fixed feature dimension, the "truncation and zero-padding" operation can be performed on the filtered ultrasonic time-domain signal. Taking the model as a one-dimensional convolutional neural network (1D-CNN) as an example, the standard length is set: According to the theoretical maximum travel time of ultrasonic waves within the detection depth range and the sampling rate, a unified time window length N is set (in this embodiment, N = 2048 sampling points). Detect the actual length L of the currently collected signal. If L > N, perform the truncation operation, that is, retain the first N data points of the signal and discard the redundant data at the end to remove long-delay reverberation noise; if L < N, perform the zero-padding operation at the end, that is, append N data points with a value of 0 at the end of the signal to expand its length to N. In this way, it ensures that the sample matrix dimensions of all inputs to the deep learning model are consistent.

[0062] After obtaining and preprocessing the ultrasonic time-domain signal, utilize the powerful non-linear feature extraction ability of the deep learning model to establish an accurate mapping relationship from the ultrasonic waveform to the corrosion rate. Aiming at the difficulties of strong scattering of internal aggregates in concrete and weak echo signals, the constructed corrosion rate prediction model can specifically be a one-dimensional convolutional neural network integrating a dual-domain attention mechanism.

[0063] The corrosion rate prediction model includes: An input layer for receiving the ultrasonic time-domain signal; A feature extraction module for extracting the local echo features of the ultrasonic time-domain signal; An attention mechanism module for adaptively weighting the local echo features to enhance the signal features related to the reflection at the steel bar interface and suppress the background noise features caused by the scattering of concrete aggregates; A regression prediction module for mapping the adaptively weighted local echo features to the detection result of the steel bar corrosion rate.

[0064] Among them, the input layer receives a one-dimensional ultrasonic time-domain sequence with a length of N after being standardized X ∈ R N×1 ; The feature extraction module, as a deep feature extractor, can be composed of multiple stacked "convolution-pooling blocks". Each layer contains a one-dimensional convolutional layer, a batch normalization layer, and a non-linear activation function for gradually extracting the local features of the signal at different time scales. The attention mechanism module can be embedded between the one-dimensional convolutional layers or after the feature extraction module to adaptively weight the feature map and suppress invalid background noise. The regression prediction module contains a global average pooling layer and a fully connected layer, and finally outputs a continuous scalar value y , which is the predicted steel bar corrosion rate.

[0065] In the feature extraction module, the convolutional layer serves as the core of feature extraction. Unlike two-dimensional convolution in image processing, this invention uses a one-dimensional convolutional kernel that slides across the temporal waveform. Assuming... l The input feature map of the layer is Then the first l Layer j The output of each convolution kernel It can be represented as: In the formula: This represents a one-dimensional convolution operation; For the first l Layer connection i The input channel and the first j The convolutional kernel weights for each output channel; For bias terms; It is the ReLU activation function, used to introduce nonlinear characteristics and enhance the model's ability to fit complex scattering signals.

[0066] See Figure 7 In order to accurately locate weak steel bar echoes in a strong noise background, this invention introduces a mechanism that combines channel attention and temporal attention.

[0067] Channel Attention Branch: Filtering "Effective Frequencies" This branch identifies which convolutional channels extract useful rebar reflection features and which channels contain only aggregate noise. Its calculation formula is: In the formula: F is the input feature map; and These represent global average pooling and max pooling along the time dimension, respectively. It is a multilayer perceptron used to learn the dependencies between channels; It is a Sigmoid activation function with output weights in the range [0, 1].

[0068] Temporal attention branch: Locking onto the "echo moment". The core function of this branch is to construct a "depth gating" mechanism to further filter out residual structural noise. In reinforced concrete members, transverse and longitudinal reinforcing bars are typically located at different depth layers, such as the thickness of the outer reinforcing bar cover. d 1 and inner layer reinforcement depth d 2. This means that the target signal and the interfering signal have a significant time-of-flight (TOF) difference in the time domain.

[0069] The corrosion rate prediction model of this invention, through training, can automatically learn a specific time window corresponding to the target rebar (based on infrared-guided estimated depth). The model assigns high weights to waveform features within this window, while signals outside the window—i.e., echoes from rebars at different depths or sidelobe scattering waves from adjacent rebars—are treated as invalid background noise and suppressed, thereby purifying the corrosion characteristics of a single target rebar. The calculation formula is as follows: In the formula: and This represents average pooling and max pooling along the channel dimension; This indicates a 7-dimensional convolution operation with a kernel size of 7, used to extract local temporal context information.

[0070] Adaptive weighted fusion: the final feature map F ' It is the result of multiplying the original feature F by the attention weight: Through this mechanism, the model can automatically "focus" on the reflected signal at the steel reinforcement interface, significantly improving its robustness in heterogeneous concrete environments.

[0071] During model training, the goal is to minimize the difference between the predicted corrosion rate and the actual corrosion rate. The mean squared error loss function can be used as the objective function. In the formula: B This refers to the training batch size; The label for the true corrosion rate of the i-th sample; The model predicts the corrosion rate; This is an L2 regularization term used to prevent model fitting and enhance generalization ability.

[0072] For example, the model training strategy can be adopted in the following ways: Optimizer: The Adam optimization algorithm is used, which utilizes its adaptive learning rate to accelerate convergence.

[0073] Learning rate decay: Set the initial learning rate to 1×10. -3 Furthermore, a cosine annealing strategy is employed to dynamically adjust the learning rate, thereby avoiding getting trapped in local optima.

[0074] Training environment: The model is built based on the TensorFlow or PyTorch deep learning framework and trained end-to-end in a GPU-accelerated environment.

[0075] In one embodiment, the above method further includes: The results of the steel corrosion rate test are associated with and stored in physical space coordinates.

[0076] After the corrosion rate prediction model completes inference and outputs the steel corrosion rate detection results at each measuring point on the target concrete structure surface, these results can be associated and stored with the physical spatial coordinates of the corresponding measuring points. Specifically, the physical spatial coordinates of each measuring point on the target concrete structure surface are bound one-to-one with the predicted steel corrosion rate value at that coordinate location, forming a structured data record containing both spatial location and corrosion degree information. By summarizing the data entries from all measuring points, a complete structural health record of the target concrete structure can be constructed, providing data support for subsequent structural condition tracing, corrosion development trend analysis, and engineering acceptance evaluation.

[0077] In one embodiment, the above method further includes: Based on the steel corrosion rate detection results and physical spatial coordinates, a corrosion rate distribution map is generated in real time on a visualization interface.

[0078] After obtaining the steel corrosion rate test results and their corresponding physical coordinates at each measuring point, this correlated data can be further used to generate a corrosion rate distribution map of the structural surface in real time on a visualization interface. Specifically, based on the physical coordinates of each measuring point, the steel corrosion rate value at that point is mapped to the corresponding position on the target concrete structural surface, and the corrosion rate is converted into intuitive color levels or grayscale changes using a preset color mapping rule, thus forming a characteristic image on the display terminal that reflects the spatial distribution of the overall corrosion degree of the structure. By observing this corrosion rate distribution map, inspectors can intuitively and quickly grasp the spatial range and severity of corrosion damage within the measured area, identify key areas of concern with high corrosion rates, and provide an intuitive decision-making basis for subsequently developing targeted repair or reinforcement plans.

[0079] In one embodiment, the above method further includes: Based on the preset corrosion rate grading threshold, the corrosion damage level of each corrosion area in the steel reinforcement corrosion rate detection results is determined, and the corresponding level of early warning is triggered.

[0080] After obtaining the steel reinforcement corrosion rate test results at each measuring point, the corrosion damage level of each corroded area can be determined based on a preset corrosion rate grading threshold, triggering corresponding level of early warning. The preset corrosion rate grading threshold can be set according to relevant civil engineering codes or structural safety assessment standards, classifying the degree of steel reinforcement corrosion into several levels. For different corrosion damage levels, corresponding level of early warning is triggered, such as by displaying different colored labels on a display terminal for differentiation, or by issuing audible and visual alarm signals to remind inspectors to pay close attention. Through this graded early warning mechanism based on quantitative test results, severely corroded and urgently requiring repair and reinforcement hazardous areas can be quickly identified during the inspection of large infrastructure projects, providing immediate risk warnings and scientific basis for engineering maintenance decisions.

[0081] For example, the grading standard for steel bar corrosion damage can be found in Table 2. If the corrosion rate is less than 5%, a green label (healthy) is displayed; if the corrosion rate is greater than 10%, a red label (severe) is displayed, and an audible and visual alarm is triggered, prompting the inspection personnel to immediately reinforce the structure. After the inspection is completed, all measurement point data can be summarized to generate an electronic spreadsheet or inspection report containing coordinates and corrosion rates for subsequent project acceptance.

[0082] Table 2

[0083] Corresponding to the aforementioned application function implementation method embodiments, the present invention also provides a steel reinforcement corrosion detection system based on passive infrared guidance and ultrasonic quantification, and corresponding embodiments.

[0084] Please see Figure 8 , Figure 8 This is a schematic diagram of the module structure of a steel corrosion detection system based on passive infrared guidance and ultrasonic quantification.

[0085] A steel reinforcement corrosion detection system based on passive infrared guidance and ultrasonic quantitative analysis includes: Infrared image acquisition unit 81 is used to acquire the original infrared thermal image sequence of the target concrete structure surface acquired by a passive infrared thermal imager. The thermal anomaly identification and localization unit 82 is used to identify thermal anomaly areas caused by steel corrosion in the infrared thermal image sequence and determine the physical spatial coordinates of the thermal anomaly areas on the surface of the target concrete structure. The ultrasonic signal acquisition unit 83 is used to acquire the ultrasonic time-domain signal of the target concrete structure surface acquired by the ultrasonic detection component at the physical space coordinates. The corrosion rate prediction unit 84 is used to input the ultrasonic time-domain signal into the pre-trained corrosion rate prediction model to obtain the detection result of the steel corrosion rate on the surface of the target concrete structure.

[0086] In one embodiment, when identifying thermal anomaly areas caused by steel corrosion in an infrared thermal image sequence, the thermal anomaly identification and localization unit 82 is specifically used for: Candidate regions exhibiting abnormal temperature gradients in infrared thermal imaging sequences were selected. Candidate areas with geometric features that are strip-shaped and whose distribution direction is consistent with the predicted direction of the steel reinforcement arrangement are identified as thermal anomaly areas caused by steel reinforcement corrosion.

[0087] In one embodiment, when determining the physical spatial coordinates of the thermal anomaly region on the surface of the target concrete structure, the thermal anomaly identification and location unit 82 is specifically used for: Determine the pixel coordinates of the thermal anomaly region in the corresponding image of the infrared thermographic sequence; Based on preset spatial resolution calibration parameters, pixel coordinates are converted into physical spatial coordinates of the target concrete structure surface.

[0088] In one embodiment, in inputting the ultrasonic time-domain signal into a pre-trained corrosion rate prediction model, the corrosion rate prediction unit 84 is specifically used for: The ultrasonic time-domain signal is preprocessed; the preprocessing includes bandpass filtering, amplitude normalization, and data length standardization. The preprocessed ultrasonic time-domain signal is input into the pre-trained corrosion rate prediction model.

[0089] In one embodiment, the detection system further includes: The storage unit is used to associate and store the steel corrosion rate detection results with the corresponding physical space coordinates.

[0090] In one embodiment, the detection system further includes: The visualization unit is used to render and generate a corrosion rate distribution map in real time on the visualization interface based on the steel corrosion rate detection results and physical spatial coordinates.

[0091] In one embodiment, the detection system further includes: The graded warning unit is used to determine the corrosion damage level of each corrosion area in the steel reinforcement corrosion rate detection results based on the preset corrosion rate grading threshold, and trigger the corresponding level of warning.

[0092] Regarding the system in the above embodiments, the specific manner in which each unit module performs operations has been described in detail in the embodiments related to the method, and will not be elaborated further here.

[0093] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for detecting steel corrosion based on passive infrared guidance and ultrasonic quantitative analysis, characterized in that, include: Obtain the original infrared thermal image sequence of the target concrete structure surface acquired by a passive infrared thermal imager; Identify the thermal anomaly region caused by steel corrosion in the infrared thermal image sequence, and determine the physical spatial coordinates of the thermal anomaly region on the surface of the target concrete structure. Acquire the ultrasonic time-domain signal of the target concrete structure surface at the physical space coordinates by the ultrasonic detection component; The ultrasonic time-domain signal is input into a pre-trained corrosion rate prediction model to obtain the steel corrosion rate detection result of the target concrete structure surface.

2. The method for detecting steel reinforcement corrosion according to claim 1, characterized in that, Identifying thermal anomaly regions caused by steel corrosion in the infrared thermal image sequence includes: Candidate regions exhibiting abnormal temperature gradients in the infrared thermal imaging sequence were selected. Candidate areas with geometric features that are strip-shaped and whose distribution direction is consistent with the predicted direction of the steel reinforcement arrangement are identified as thermal anomaly areas caused by steel reinforcement corrosion.

3. The method for detecting steel reinforcement corrosion according to claim 1, characterized in that, Determining the physical spatial coordinates of the thermal anomaly region on the surface of the target concrete structure includes: Determine the pixel coordinates of the thermal anomaly region in the corresponding image of the infrared thermographic sequence; Based on preset spatial resolution calibration parameters, the pixel coordinates are converted into physical spatial coordinates of the target concrete structure surface.

4. The method for detecting steel reinforcement corrosion according to claim 1, characterized in that, The corrosion rate prediction model includes: The input layer is used to receive the ultrasonic time-domain signal; The feature extraction module is used to extract the local echo features of the ultrasonic time-domain signal; An attention mechanism module is used to adaptively weight the local echo features to enhance the signal features related to the reflection from the steel reinforcement interface and suppress the background noise features caused by the scattering of concrete aggregate. The regression prediction module is used to map adaptively weighted local echo features to steel corrosion rate detection results.

5. The method for detecting steel reinforcement corrosion according to claim 1, characterized in that, The method further includes: The steel reinforcement corrosion rate detection results are associated and stored with the corresponding physical space coordinates.

6. The method for detecting steel reinforcement corrosion according to claim 1, characterized in that, The method further includes: Based on the steel corrosion rate detection results and physical space coordinates, a corrosion rate distribution map is generated in real time on the visualization interface.

7. The method for detecting steel reinforcement corrosion according to claim 1, characterized in that, The method further includes: Based on the preset corrosion rate grading threshold, the corrosion damage level of each corrosion area in the steel bar corrosion rate detection results is determined, and a corresponding level of early warning is triggered.

8. The method for detecting steel reinforcement corrosion according to claim 1, characterized in that, The ultrasonic time-domain signal is input into a pre-trained corrosion rate prediction model, including: The ultrasonic time-domain signal is preprocessed; the preprocessing includes bandpass filtering, amplitude normalization, and data length standardization. The preprocessed ultrasonic time-domain signal is input into the pre-trained corrosion rate prediction model.

9. A steel reinforcement corrosion detection system based on passive infrared guidance and ultrasonic quantitative analysis, characterized in that, include: The infrared image acquisition unit is used to acquire the original infrared thermal image sequence of the target concrete structure surface acquired by the passive infrared thermal imager. A thermal anomaly identification and localization unit is used to identify thermal anomaly areas caused by steel corrosion in the infrared thermal image sequence and determine the physical spatial coordinates of the thermal anomaly areas on the surface of the target concrete structure. An ultrasonic signal acquisition unit is used to acquire ultrasonic time-domain signals of the target concrete structure surface collected by the ultrasonic detection component at the physical space coordinates. The corrosion rate prediction unit is used to input the ultrasonic time-domain signal into a pre-trained corrosion rate prediction model to obtain the detection result of the steel corrosion rate on the surface of the target concrete structure.

Images