Steel box girder hidden damage identification method based on magnetic memory effect

Abstract

The invention discloses a steel box girder hidden damage identification method based on a magnetic memory effect, and relates to the technical field of bridge structure health monitoring, and the method comprises the steps: employing a magnetic memory detector, arranging a probe to detect the normal component of a leakage magnetic field on the surface of a steel box girder, carrying out the normalization processing of an original signal in order to eliminate the influence of an initial residual magnetic field, calculating to obtain a benchmark magnetic signal variable quantity; detecting lines are arranged on the upper flange, the lower flange and the web of the steel box girder at the interval of 5mm, and detecting points are arranged on the detecting lines at the interval of 1mm; carrying out graded loading on the steel box girder, synchronously acquiring strain data and magnetic memory signal data, and constructing a structured data set; a magnetic signal variation distribution curve is drawn according to the data set, the curve form is analyzed to recognize a local extreme point, and the position of the extreme point is a local buckling damage area; and verifying the consistency of the position of the extreme point and the actual buckling damage area by comparing the final damage form of the test piece, so as to realize high-precision positioning of the hidden damage.

Description

Technical Field

[0001] This invention relates to the field of bridge structural health monitoring technology, and in particular to a method for identifying hidden damage in steel box girders based on magnetic memory effect. Background Technology

[0002] As the core load-bearing structure of long-span bridges, the stability of steel box girders directly affects the overall safety of the bridge. However, under the coupled effect of complex alternating loads and welding residual stress, the compression areas such as the steel box girder walls and stiffeners are prone to local buckling damage. This type of damage initially manifests as micro-buckling deformation, with no obvious cracking characteristics, but it leads to a sharp drop in structural stiffness and stress redistribution. Currently, non-destructive testing for local buckling of steel box girders mainly relies on strain gauge monitoring, ultrasonic thickness measurement, and three-dimensional laser scanning technology, but these technologies have significant limitations: strain gauges are not sensitive enough to micron-level wrinkling deformation in the early stage of buckling; ultrasonic testing relies on thickness changes and cannot identify damage in the early stage where the wall thickness has not been significantly reduced; existing technologies generally have the limitation of making it difficult to establish a quantitative relationship between buckling damage evolution and mechanical state, and cannot accurately quantify and assess micro-damage. Most importantly, they cannot effectively capture and warn of critical stress states, that is, the critical point when the structure is about to buckle instability or damage accelerates, which makes it impossible to actively intervene before catastrophic failure, becoming a major blind spot in the safety assurance system of steel box girders. Therefore, there is an urgent need for a high-precision, full-cycle damage identification method that can simultaneously achieve early location of hidden damage, real-time early warning of critical stress state, and quantitative assessment of damage degree, in order to fill the gap in the existing technology in key early warning and quantitative assessment capabilities, and provide more reliable protection for the safe operation of bridge structures. Summary of the Invention

[0003] This application provides a method for identifying hidden damage in steel box girders based on magnetic memory effect, which solves the problems of inaccurate early location of hidden damage, insufficient early warning capability of critical stress state, and difficulty in quantitative assessment of damage degree in the prior art. It achieves the technical effects of high-precision location of local buckling damage area, early warning of critical yield state, and quantitative assessment of web bending and shear bearing capacity.

[0004] This application provides a method for identifying hidden damage in steel box girders based on magnetic memory effect, including: S1: A magnetic memory detector, equipped with a probe, is used to detect the normal component of the leakage magnetic field on the surface of the steel box girder. To eliminate the influence of the initial residual magnetic field, the original signal is normalized, and the baseline magnetic signal change is calculated. , in, To reflect only the net change in magnetic signal caused by load variation, Let be the magnetic signal value under the i-th level load. The zero-load signal value; S2: Detection lines are arranged at 5mm intervals on the upper flange, lower flange, and web of the steel box girder, and detection points are set at 1mm intervals on each detection line; the steel box girder is subjected to graded loading, and strain data and magnetic memory signal data are collected simultaneously to construct a structured dataset; S3: Plot the distribution curve of magnetic signal change based on the dataset, analyze the curve shape to identify local extreme points, and the location of the extreme points is the local buckling damage area; by comparing the final failure morphology of the specimen, verify the consistency between the extreme point location and the actual buckling damage area, and achieve high-precision positioning of hidden damage.

[0005] Furthermore, the graded loading includes: applying loads step by step with a preset load gradient, the load gradient being set based on the yield load of the steel box girder; maintaining load stability after each load level to ensure that the structural response tends to stabilize and completing the synchronous acquisition of strain data and magnetic memory signal data at all detection points under that load level; and continuously loading until a predetermined termination load condition is reached, the termination load condition including the steel box girder entering the yield stage, the occurrence of macroscopic buckling deformation, and reaching the predetermined maximum loading load.

[0006] Furthermore, the original signal also includes: defining magnetic parameters and integral area parameters based on the original signal; The formula for calculating the magnetic parameters is: , in, For the average value parameter, The total number of testing sites. Let j be the measured value of the normal component of the magnetic signal at the j-th detection point under a specific load. To calculate the arithmetic mean of the magnetic signal values ​​at all detection points; The formula for calculating the integral area parameter is as follows: , in, For the integral area parameter, and These are the upper and lower limits of the integral. This represents the length increment along the detection line direction. for Along the detection line from arrive Integrate and calculate the area under the curve.

[0007] Furthermore, the magnetic parameters include plotting curves showing the magnetic parameters as a function of load, and analyzing the trend of the curves: The elastic phase curve rises as the load increases; The appearance of a peak point on the yield curve indicates the onset of yielding; The critical stage curve shows a reversal point in the trend at the critical stress state. This reversal point serves as a damage warning sign that the steel box girder has reached the critical stress state. The trend reversal point of the curve refers to the inflection point where the average parameter and the load curve change from rising to falling and from falling to rising.

[0008] Furthermore, the curve trend reversal point also includes: defining key magnetic parameters based on the location of local extreme points; The key magnetic parameters are calculated based on the changes in the normal component of the magnetic memory signal, including: Parameter: The area enclosed by the detection line and the coordinate axis With the length of the detection line The ratio; Parameter: The average value of the absolute value of the magnetic signal on the detection line; Parameter: The maximum absolute value of the magnetic gradient on the detection line; Parameter: The average absolute value of the magnetic gradient on the detection line; Establish a quantitative relationship model between key magnetic parameters and stress at local extreme points.

[0009] Furthermore, the method also includes web bending and shear capacity assessment: defining the absolute value parameter of the average magnetic signal in the web region: , in, The absolute value parameter of the average magnetic signal. The total number of testing sites. Let be the absolute value of the magnetic signal at the i-th detection point; Establishing dimensionless parameters based on the absolute value parameters of the average magnetic signal: , in, For dimensionless parameters, The absolute value parameter of the average magnetic signal in the initial state. This is the absolute value parameter of the average magnetic signal in the web region under the current load; A linear model is constructed by linearly fitting the dimensionless parameters to the load ratio, which is used to evaluate the bending and shear bearing capacity of the web. The formula for calculating the load ratio is: ,in, For the current load, This is the yield load; The correlation coefficient of the linear model A value greater than 0.85 ensures the reliability of the evaluation results.

[0010] Furthermore, the method also includes multimodal data fusion: deploying acoustic emission sensors, infrared thermal imagers, and fiber Bragg grating sensors to synchronously acquire data with a magnetic memory detector; the deployed acoustic emission sensors are used to capture acoustic signals of microcrack propagation, the infrared thermal imagers are used to monitor temperature anomalies caused by stress concentration, and the fiber Bragg grating sensors are used to measure strain distribution under dynamic loads; the multimodal data are time-stamped and synchronized to construct a comprehensive damage identification dataset.

[0011] Furthermore, the fused multimodal data also includes dynamic evolution modeling: analyzing the evolution trend of damage parameters based on the comprehensive damage identification dataset, and calculating the damage index. , in, The damage index is given by time t. This represents the average value of the magnetic parameters. For acoustic emission energy, For the rate of temperature rise, For strain energy density, The corresponding weighting coefficients were determined through calibration using experimental data, and ; A dynamic damage evolution model is established to describe the damage state transition process, including stable, nascent, and extended states; through the evolution law of the state, early warning of critical yield state and prediction of damage extension are achieved.

[0012] Furthermore, the calculation of the damage index also includes uncertainty quantification: defining uncertainty parameters for each modal feature in the multimodal data, the uncertainty parameters including the measurement error range of the magnetic memory signal, the signal-to-noise ratio of the acoustic emission signal, the temperature error band of the thermal imaging data, and the accuracy error of the strain data; and outputting the mean, variance, and confidence interval of the damage index through a Monte Carlo simulation model.

[0013] Furthermore, the method also includes calculating the overall confidence level: , in, To assess the overall confidence level, For the average uncertainty, As a modal consistency index, The root mean square error between the damage dynamic evolution model prediction results and historical monitoring data is given. The corresponding weighting coefficients are determined through calibration using historical data and satisfy the following conditions: .

[0014] One or more technical solutions provided in this application have at least the following technical effects or advantages: By employing magnetic memory detection technology, high-precision positioning of hidden damage in steel box girders was achieved, enabling early warning of critical stress states and quantitative assessment of damage levels. Dynamic evolution modeling and uncertainty quantification improved the robustness and reliability of damage identification, while a comprehensive confidence mechanism ensured the accuracy of the warning results. Attached Figure Description

[0015] Figure 1 This is a flowchart of a method for identifying hidden damage in steel box girders based on magnetic memory effect in an embodiment of the present invention; Figure 2 This is a schematic diagram of the normal component distribution curve of the specimen without butt weld in an embodiment of the present invention; Figure 3 This is a schematic diagram of the normal component distribution curve of the specimen containing the butt weld in an embodiment of the present invention; Figure 4 This is a schematic diagram of the magnetic parameter average value-load relationship curve and strain-load relationship curve of the specimen without butt weld in the embodiment of the present invention; Figure 5 This is a schematic diagram of the magnetic parameter average value-load relationship curve and strain-load relationship curve of the specimen containing butt weld in the embodiment of the present invention; Figure 6 This is a curve showing the relationship between characteristic magnetic parameters and stress at a local buckling location in an embodiment of the present invention; Figure 7 This is a schematic diagram of the relationship between the web region parameter m and the load ratio in an embodiment of the present invention. Detailed Implementation

[0016] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0018] Example 1: As Figure 1 As shown, a method for identifying hidden damage in steel box girders based on magnetic memory effect is presented.

[0019] S1: A magnetic memory detector is used, equipped with a probe to detect the normal component of the leakage magnetic field on the surface of the steel box girder. In order to eliminate the influence of the initial residual magnetic field, the original signal is normalized and the change in the benchmark magnetic signal is calculated. Specifically, a high-precision magnetic memory detector (such as the EMS-2003 model) equipped with a dual-channel pen probe is used to detect the normal component of the leakage magnetic field on the surface of the steel box girder. The equipment performance must meet the following requirements: resolution of not less than 1 A / m to ensure the capture of weak magnetic signal changes; and ranging accuracy of not less than 1 mm to ensure the accuracy of the detection point positioning.

[0020] To eliminate the influence of the initial residual magnetic field of the steel box girder and obtain the net change in magnetic signal caused only by load variation, the specimen needs to be normalized before measurement. The normalized magnetic signal change is calculated using the following formula: , in, This represents the normalized change in magnetic signal, reflecting the net change in magnetic signal caused by load variations. Let be the normal component value of the magnetic signal acquired under the i-th level load. The reference value for the normal component of the magnetic signal under zero load (i.e., initial state).

[0021] S2: Detection lines are arranged at 5mm intervals on the upper flange, lower flange, and web of the steel box girder, and detection points are set at 1mm intervals on each detection line; the steel box girder is subjected to graded loading, and strain data and magnetic memory signal data are collected simultaneously to construct a structured dataset; Specifically, in the key areas of the steel box girder: the surfaces of the upper flange, lower flange, and web are uniformly arranged with a test line spacing of 5 mm to ensure coverage of high-stress areas; dense test points are set at 1 mm intervals on each test line to achieve high-resolution capture of microscopic magnetic signal changes. The arrangement of test points must ensure uniformity and repeatability to eliminate measurement errors.

[0022] The graded loading includes: applying loads in stages according to a preset load gradient, the load gradient being set based on the yield load of the steel box girder; maintaining load stability after each load stage to ensure that the structural response tends to stabilize and completing the synchronous acquisition of strain data and magnetic memory signal data of all detection points under that load stage; and continuously loading until a predetermined termination load condition is reached, the termination load condition including the steel box girder entering the yield stage, the occurrence of macroscopic buckling deformation, and reaching the predetermined maximum loading load.

[0023] Specifically, the steel box girder is subjected to graded loading, with the load gradient based on the yield load of the steel box girder. (Settings). For example, increasing the load by 10% for each load level. This simulates actual load conditions or test loads. After each load level is applied, the load is kept stable for a period of time (e.g., 1-2 minutes) to ensure the structural response tends to stabilize. Strain data and magnetic memory signal data are simultaneously collected from all detection points under that load level. Strain data is measured using connected strain gauges, and the normal component of the magnetic memory signal data is collected using a magnetic memory detector. The loading process continues until the predetermined termination load conditions are reached, including the steel box girder entering the yielding stage, the occurrence of macroscopic buckling deformation, and the reaching of the maximum load.

[0024] The collected data is organized according to the location of the detection point and the load level, including the original magnetic signal, strain value, and calculated values. , construct a structured dataset.

[0025] S3: Plot the distribution curve of magnetic signal change based on the dataset, analyze the curve shape to identify local extreme points, and the location of the extreme points is the local buckling damage area; by comparing the final failure morphology of the specimen, verify the consistency between the extreme point location and the actual buckling damage area, and achieve high-precision positioning of hidden damage.

[0026] Specifically, based on the structured dataset, a distribution curve of magnetic signal variation is plotted. For each detection line, the x-axis is the location of the detection point (unit: mm), and... The value is the ordinate (unit: A / m), plot The curve should clearly show the distribution characteristics of the normal component of the magnetic signal, for example, in a specimen without a butt weld. Figure 2 As shown, the curve may indicate a significant presence of pressure in the central compression area of ​​the specimen. Extreme points; and in specimens containing butt welds, such as Figure 3 As shown, the curve may indicate the presence of extreme points near the weld and heat-affected zone.

[0027] Extreme points include local maxima (peaks) or local minima (valleys). These points correspond to regions of abrupt changes in the magnetic field and serve as indicators of local buckling damage. The identification method employs numerical analysis algorithms, such as gradient calculation, to determine... The point on the curve where the first derivative is zero and the sign of the second derivative changes. The location of the extreme point directly corresponds to the initiation region of local buckling damage. For example, in the elastic stage, the extreme point may indicate the initiation of micro-buckling.

[0028] After the graded loading test, the actual failure mode of the steel box girder specimens was observed, and the local buckling damage area was recorded. The locations of extreme points were spatially compared with the actual damage areas. The verification results show that the locations of extreme points are highly consistent with the actual buckling damage areas. For example, in specimens without butt welds, the extreme points are concentrated in the compression zone in the middle of the specimen, which coincides with the actual wrinkle location; in specimens with butt welds, the extreme points are concentrated at the weld ends, corresponding to the heat-affected zone damage. This comparison verifies the sensitivity of magnetic memory signals to hidden damage, and the positioning error can be controlled within ±5mm, achieving high-precision positioning.

[0029] The original signal also includes: defining magnetic parameters and integral area parameters based on the original signal; Specifically, magnetic parameters and integral area parameters are key indicators based on the variation of the normal component of the magnetic memory signal, used to quantify the stress state and damage evolution of the steel box girder. The formula for calculating magnetic parameters is as follows: , in, The average value parameter characterizes the change in overall magnetization intensity within the detection area. The total number of testing sites. Let j be the measured value of the normal component of the magnetic signal at the j-th detection point under a specific load. To calculate the arithmetic mean of the magnetic signal values ​​at all detection points; The formula for calculating the integral area parameter is: , in, This is the integral area parameter, reflecting the degree of accumulation in the stress concentration region. and These are the upper and lower limits of integration, corresponding to the start and end points of the detection line. This represents the length increment along the detection line direction. for Along the detection line from arrive Integrate the curve and calculate the area under the curve; the larger the value, the more significant the stress concentration.

[0030] The magnetic parameters include plotting curves showing the changes in magnetic parameters with load and analyzing the trend of the curves; Specifically, based on magnetic parameters The curve of load F (i.e.) The -F curve can be used to analyze the following characteristic stages: Elastic stage: The curve rises with increasing load, reflecting the increased magnetization caused by the reversible movement of magnetic domains. Yield stage: When the load approaches the yield load... (approximately 90%) When the elastic-plastic transition point (critical stress state) is reached, the curve shows a peak, indicating that the material is beginning to yield. Critical stage: At the elastic-plastic transition point (critical stress state), the curve shows a trend reversal point, a damage warning sign that the steel box girder has reached the critical stress state. For example, as... Figure 4 As shown, the specimen without butt welds The -F curve clearly shows the rise in the elastic stage, the peak value near the yield point, and the trend reversal point under higher loads (critical state). The ε-F curve shows typical elastoplastic behavior. Figure 5 As shown, the specimen includes a butt weld. -F curve and specimen without butt weld The -F curve is similar, but the peak and reversal point may appear earlier, reflecting the influence of the weld on the load-bearing capacity and critical state.

[0031] The curve trend reversal point also includes: defining key magnetic parameters based on the location of local extreme points; Specifically, the key magnetic parameters are calculated based on the normal component of the magnetic memory signal, and each parameter reflects different magnetic field characteristics. The area enclosed by the detection line and the coordinate axes With the length of the detection line The ratio is used to eliminate the influence of the detection line length and highlight the difference in magnetization caused by stress. The calculation formula is: , in, for Along the detection line from arrive The integral area (i.e.) L is the length of the detection line. The larger the value, the more significant the stress concentration, which is positively correlated with the degree of local buckling damage.

[0032] The average absolute value of the magnetic signal detected online is used as a normalization benchmark to reflect the overall change in magnetic field strength. The calculation formula is as follows: , in, The total number of testing sites. It is the change in magnetic signal at the j-th detection point, used to compare the damage evolution under different loads.

[0033] To detect the maximum absolute value of the magnetic gradient on the line, reflecting the degree of drastic change in the local magnetic field, it is used to identify stress concentration peak points. Magnetic gradient Calculated as The first derivative with respect to position (i.e.) Take the maximum value on the detection line: , high The value corresponds to the micro-buckling initiation area and is a sensitive indicator of early damage.

[0034] The average absolute value of the magnetic gradient on the detection line represents the overall magnetic gradient level and is used to assess the uniformity of stress distribution. The calculation formula is: , Related to the average stress level, an increase in the value indicates damage propagation.

[0035] Establish a quantitative relationship model between key magnetic parameters and stress at local extreme points; for example, Figure 6 As shown, the area of ​​magnetic parameter integration ( ) and mean stress ( The relationship between magnetization M and stress σ and magnetic field H is shown in the curve, exhibiting a typical three-stage trend of decreasing-increasing-decreasing. In the initial stage, the initial stress induces reversible movement of the domain walls. If the stress direction is inconsistent with the magnetization direction of some domains, it leads to local magnetic moment rearrangement, resulting in a temporary demagnetization effect. As stress increases, the domains gradually align in favorable directions, and the magnetization recovers and strengthens. The peak and decline stages can be qualitatively explained using the extended Jiles-Atherton-Sablik magnetomechanical effect model. This model describes the relationship between magnetization M and stress σ and magnetic field H. Under a constant geomagnetic field, the model predicts that under tensile or compressive stress, the magnetization will first increase and then decrease with increasing stress, consistent with observed... The peak value and subsequent decline are qualitatively consistent. The signal reduction mechanism during the plastic stage is as follows: Unlike the elastic stage, plastic deformation leads to a large accumulation of dislocations. High-density dislocations act as strong pinning points for domain wall motion, hindering it. Under a weak external field, the domain walls struggle to overcome these pinning points, resulting in impaired magnetization in the dislocation accumulation region, reduced permeability, and macroscopically manifested as a decrease in the magnetic memory signal. A polynomial fitting method was used to establish a quantitative relationship, obtaining a quantitative model with a high correlation coefficient (R²>0.85), which was used to invert the stress at this location.

[0036] The method also includes web bending and shear capacity assessment: defining the average absolute value parameter of the magnetic signal in the web region. , in, The absolute value parameter of the average magnetic signal. The total number of testing sites. Let be the absolute value of the magnetic signal at the i-th detection point; Establishing dimensionless parameters based on the absolute value parameters of the average magnetic signal: , in, For dimensionless parameters, The absolute value parameter of the average magnetic signal in the initial state. This is the absolute value parameter of the average magnetic signal in the web region under the current load; Specifically, dimensionless parameters Ratio to load Perform linear fitting and establish a model: , in, For the current load, For yield load, and The fitting coefficients, calibrated using experimental data, show a correlation coefficient R² greater than 0.85, ensuring the reliability of the assessment. The model can be used for real-time back-calculation of load ratios. , like Figure 7 As shown, dynamic monitoring of the web bending and shear bearing capacity is achieved, with the error controlled within ±5%.

[0037] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: This application employs a method for identifying hidden damage in steel box girders based on the magnetic memory effect. By detecting changes in the normal component of the leakage magnetic field on the surface of the steel box girder, it achieves high-precision localization of hidden damage, early warning of critical stress states, and quantitative assessment of damage severity. The method eliminates the influence of initial residual magnetic fields through magnetic signal acquisition and preprocessing. A structured dataset is constructed by synchronously acquiring data using dense detection lines and graded loading. Local extreme points are identified based on the magnetic signal distribution curve, enabling precise localization of hidden stress concentration areas with an error controlled within ±5mm. Analysis of the curve characteristics of magnetic parameters and integral area parameters changing with load reveals an increase in the elastic stage, a peak at yielding, and a trend reversal point at the critical state. A force-magnetic quantification model at local buckling points and a web bending-shear bearing capacity assessment model are established to achieve real-time assessment of damage severity.

[0038] Example 2: Example 1 achieved high-precision positioning of hidden damage in steel box girders based on magnetic memory effect, early warning of critical stress state, and quantitative assessment of damage degree. However, it still has the problem that relying on a single magnetic memory data may lead to insufficient reliability under complex working conditions. This example further supplements Example 1.

[0039] The method further includes multimodal data fusion: deploying acoustic emission sensors, infrared thermal imagers, and fiber Bragg grating sensors to synchronously acquire data with a magnetic memory detector; the deployed acoustic emission sensors are used to capture acoustic signals of microcrack propagation, the infrared thermal imagers are used to monitor temperature anomalies caused by stress concentration, and the fiber Bragg grating sensors are used to measure strain distribution under dynamic loads; the multimodal data are time-stamped and synchronized to construct a comprehensive damage identification dataset.

[0040] Specifically, multiple sensors are deployed in conjunction with a magnetic memory detector to achieve data complementarity and cross-validation. Acoustic emission sensors are deployed in key areas of the steel box girder, such as stress concentration zones, to capture acoustic signals generated by microcrack propagation and plastic deformation. The characteristic parameters of the acoustic emission signals can reflect the initiation of damage, have high sensitivity, and can identify microcracks at an early stage. Infrared thermal imagers use non-contact scanning to monitor the surface temperature distribution of the steel box girder. Stress concentration leads to internal frictional heat or plastic heat dissipation, causing local temperature rises. Areas with abnormal temperatures correspond to potential damage zones. The thermal imager has a resolution of no less than 0.1°C and a sampling frequency of ≥5Hz to ensure real-time capture of temperature changes under dynamic loads. Fiber optic grating arrays are deployed along the detection line, with each grating node measuring local strain values ​​for high-frequency strain acquisition under dynamic loads, providing spatiotemporal evolution data of strain distribution. The magnetic memory detector simultaneously acquires the normal component of the leakage magnetic field, forming a multi-physics monitoring network with the sensors.

[0041] Linear interpolation was used to resample data from acoustic emission sensors, infrared thermal imagers, and fiber Bragg grating sensors to a unified time series, with synchronization errors controlled within ±0.5ms to ensure time alignment of multimodal data. The raw multimodal data underwent filtering, normalization, and coordinate mapping preprocessing. The positions of each sensor were calibrated using 3D laser scanning, mapping the acoustic emission array nodes, thermal imager pixel coordinates, and fiber Bragg grating detection points to the same coordinate system on the steel box girder surface, aligning with the magnetic memory detection lines to achieve spatial data fusion. The synchronized multimodal data was then integrated into a structured dataset based on time series and spatial location. The dataset includes multidimensional features such as magnetic signals, acoustic emission energy, temperature, and strain. Each data record corresponds to a specific time point, load level, and detection point location, supporting correlation analysis.

[0042] The fused multimodal data also includes dynamic evolution modeling: analyzing the evolution trend of damage parameters based on the comprehensive damage identification dataset and calculating the damage index. , in, The damage index is given by time t. This represents the average value of the magnetic parameters. For acoustic emission energy, For the rate of temperature rise, For strain energy density, The corresponding weighting coefficients were determined through calibration using experimental data, and ; A dynamic damage evolution model is established to describe the damage state transition process, including stable, nascent, and extended states; through the evolution law of the state, early warning of critical yield state and prediction of damage extension are achieved.

[0043] Specifically, The normalized average value of the magnetic parameters over time t: , in, It is the normalized change in magnetic signal at the j-th detection point at time t, reflecting the net magnetic change caused by the load; The measured value of the normal component of the magnetic signal at the j-th detection point at time t. This is the reference magnetic signal value at the j-th detection point under zero-load conditions. The total number of testing sites. It is the normalized spatial average value, which eliminates the influence of the initial residual magnetic field and is used to quantify the change of the overall magnetization intensity over time.

[0044] Normalized acoustic emission energy: , in, For acoustic emission sensors in time The amplitude of the acquired signal, For acoustic emission signal energy within a time window [ Integrals within ] For the normalized reference energy, and This represents the current time and the length of the integration window.

[0045] For the rate of temperature rise: , in, Let be the surface temperature of the steel box girder at time t. For time intervals, The normalized reference temperature is the initial temperature or the ambient temperature.

[0046] For strain energy density: , in, For stress, For strain, the strain data measured by the fiber optic grating sensor is combined with the material constitutive relation to calculate the strain. Let V be the volume of the monitoring area of ​​the steel box girder, and let V be the total strain energy within that area. Indicates the change from zero strain to the current strain The strain energy stored per unit volume during the process.

[0047] The damage dynamic evolution model divides the damage process into three states, and realizes the state transition through the evolution law of the damage index: The steady state corresponds to the lossless stage, with a low damage index, such as... <0.3, parameter changes are stable. Magnetic signal As the load elasticity increases, the acoustic emission energy and rate of temperature rise No significant peak value, strain energy density Linear growth. This state indicates structural safety, requiring no intervention.

[0048] The nascent state corresponds to the initial stage of damage development, with a moderate damage index, such as 0.3≤ <0.6, the parameter shows a sudden change. Magnetic signal. Peak values ​​appear, acoustic emission energy Sudden increase, rate of temperature rise Increase, strain energy density Non-linear growth. When When the threshold is exceeded at multiple consecutive time points and the slope increases, the system transitions from a stable state to a nascent state. This state provides an early warning, indicating that damage has begun to develop.

[0049] The extended state corresponds to the damage acceleration phase, with a high damage index, such as... ≥0.6 indicates drastic parameter changes. Magnetic signal. A trend reversal has occurred, acoustic emission energy Sustained high value, rate of temperature rise Significantly increased strain energy density Saturation or decline. When When the critical value is reached and the second derivative changes sign, the system transitions from the nascent state to the extended state. This state indicates accelerated damage and requires immediate intervention.

[0050] Accurate early warning can be achieved by understanding the laws governing state evolution; in the nascent stage, when... If the growth rate exceeds a preset threshold, such as a slope > 0.1 / min, or if multimodal parameters show synchronous anomalies, such as the magnetic peak value coinciding with the acoustic emission energy peak value, a critical yielding state warning is triggered. This warning can detect buckling signs 15%-50% of the load in advance, providing a time window for maintenance.

[0051] Damage evolution models are trained based on historical data, such as using LSTM neural networks, to predict... The future value. Through extrapolation. Curves predict damage propagation paths and remaining lifetime. For example, if the model predicts... The value will reach 0.7 at the next load level, indicating potential macroscopic buckling. Unloading or reinforcement is recommended.

[0052] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: This application employs multimodal data fusion and dynamic evolution modeling. By deploying acoustic emission sensors, infrared thermal imagers, and fiber optic grating sensors to synchronously collect data with a magnetic memory detector, a comprehensive damage identification dataset is constructed, enabling more reliable identification of hidden damage in steel box girders. The complementary nature of multimodal data overcomes the potential reliability issues of single magnetic memory data under complex working conditions. By capturing microcrack propagation through acoustic emission energy, monitoring temperature anomalies through infrared thermal imaging, and measuring strain distribution through fiber optic gratings, cross-validation is achieved with magnetic memory signals, significantly improving the robustness and accuracy of damage identification. Dynamic evolution modeling, based on the damage index analysis parameter evolution trend, enables real-time description of the damage state transition process.

[0053] Example 3: Example 2 achieved multimodal data fusion and dynamic evolution modeling, improving the reliability and accuracy of identifying hidden damage in steel box girders. However, it still has the problem of not explicitly processing data uncertainty and quantifying the confidence level of early warning results. This example further supplements Example 2.

[0054] The calculation of the damage index also includes uncertainty quantification: defining uncertainty parameters for each modal feature in the multimodal data, including the measurement error range of the magnetic memory signal, the signal-to-noise ratio of the acoustic emission signal, the temperature error band of the thermal imaging data, and the accuracy error of the strain data; and outputting the mean, variance, and confidence interval of the damage index through a Monte Carlo simulation model.

[0055] Specifically, the measurement error range of the magnetic memory signal is the absolute error range of the normal component measurement of the magnetic memory detector, calculated based on equipment specifications and actual calibration data: , in, To measure the error range, This is a conservative coefficient. , For resolution.

[0056] The signal-to-noise ratio (SNR) of an acoustic emission signal is the ratio of the acoustic emission signal power to the background noise power, reflecting signal quality. A lower SNR indicates higher uncertainty. The calculation method is as follows: , in, For signal-to-noise ratio, For signal power, The noise power is calibrated using pre-collected noise samples.

[0057] The temperature error band of thermal imaging data is the absolute error range of temperature measurement by the infrared thermal imager, calculated as follows: , in, This is the temperature error band. Due to instrument error, This is due to environmental error.

[0058] The accuracy error of the strain data is the absolute error range of the strain measurement of the fiber Bragg grating sensor, calculated as follows: , in, For accuracy error, This is half the width of the error range.

[0059] The Monte Carlo simulation model is used for iteration. For each iteration, the damage index after perturbation is calculated: , in, Let be the damage index value of time t in the k-th Monte Carlo iteration, where k is the number of Monte Carlo iterations.

[0060] right Statistical analysis of the values ​​yields the mean and variance: , in, Let be the mean of the damage index at time point t. Let be the total number of iterations in the Monte Carlo simulation, and represent the expected value of the damage index.

[0061] , in, The variance of the damage index at time t is used to quantify uncertainty.

[0062] All The values ​​are sorted in ascending order to form an ordered sequence, and the 95% confidence interval corresponds to the middle 95% of the data range.

[0063] The method also includes calculating the overall confidence level: , in, To assess the overall confidence level, To average the uncertainty, the variance of the Monte Carlo output is calculated: , For time points, The standard deviation of the damage index over time t is denoted by t. A smaller value indicates lower uncertainty and higher confidence. The modal consistency index is the average correlation coefficient between each modal parameter and the damage index. , The Pearson correlation coefficient is the coefficient of magnitude of the modality. The closer the value is to 1, the higher the modal consistency. The root mean square error between the damage dynamic evolution model prediction results and historical monitoring data: , The root mean square error, This represents the total number of historical monitoring data points. For the i-th time point, For at a certain point in time The predicted value of the damage index, For at a certain point in time The damage index observations are from historical monitoring datasets. The corresponding weighting coefficients are determined through calibration using historical data and satisfy the following conditions: .

[0064] Used to assess the reliability of damage status, under steady-state conditions. A value of ≥0.8 indicates high confidence, structural safety, and no intervention required. A value <0.6 indicates low confidence; it is recommended to increase the monitoring frequency or verify the data quality. In the nascent stage... Used to correct warning thresholds, for example, when When the threshold for triggering the nascent state is ≥0.7, it can be adjusted to: ≥0.28 (original threshold 0.3), achieving early warning; when When <0.5, the threshold is increased to ≥0.35, reducing false alarms. In extended state, Used to confirm accelerated damage. If the value is ≥0.9, an intervention alert will be triggered immediately. A value <0.6 requires verification using other metrics (such as visual inspection). (Pass / Pass) By adjusting the threshold, the system can dynamically adapt to uncertainty and improve the accuracy of early warnings.

[0065] When the damage index D(t) enters the initiation or expansion state, the system simultaneously outputs... Value. The warning signal is divided into three levels: Level 1 is a high-confidence warning (…). ≥0.8) automatically triggers the maintenance plan and sends a high-priority alarm; the second level is a medium-confidence warning (0.5≤ <0.8 indicates a need for manual review; short-term monitoring is recommended. The third level is a low-confidence warning. <0.5), only logs are recorded, no action is triggered, and data recalibration is required.

[0066] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: This application employs multimodal data uncertainty modeling and comprehensive confidence assessment. By defining precise uncertainty parameters for magnetic memory signals, acoustic emission signals, thermal imaging data, and strain data, and utilizing Monte Carlo simulation to output the mean, variance, and confidence interval of the damage index, it achieves the reliable quantification of hidden damage identification in steel box girders. By calculating the comprehensive confidence level and integrating average uncertainty, modal consistency, and prediction error, it dynamically adjusts the early warning threshold and graded early warning signals, thereby improving the accuracy and robustness of damage early warning.

[0067] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying hidden damage in steel box girders based on magnetic memory effect, characterized in that, include: S1: A magnetic memory detector, equipped with a probe, is used to detect the normal component of the leakage magnetic field on the surface of the steel box girder. To eliminate the influence of the initial residual magnetic field, the original signal is normalized, and the baseline magnetic signal change is calculated. ， in, To reflect only the net change in magnetic signal caused by load variation, Let be the magnetic signal value under the i-th level load. The zero-load signal value; S2: Detection lines are arranged at 5mm intervals on the upper flange, lower flange, and web of the steel box girder, and detection points are set at 1mm intervals on each detection line; the steel box girder is subjected to graded loading, and strain data and magnetic memory signal data are collected simultaneously to construct a structured dataset; S3: Plot the distribution curve of magnetic signal change based on the dataset, analyze the curve shape to identify local extreme points, and the location of the extreme points is the local buckling damage area; by comparing the final failure morphology of the specimen, verify the consistency between the extreme point location and the actual buckling damage area, and achieve high-precision positioning of hidden damage.

2. The method for identifying hidden damage in steel box girders based on magnetic memory effect as described in claim 1, characterized in that, The graded loading includes: applying loads in stages according to a preset load gradient, the load gradient being set based on the yield load of the steel box girder; maintaining load stability after each load stage to ensure that the structural response tends to stabilize and completing the synchronous acquisition of strain data and magnetic memory signal data of all detection points under that load stage; and continuously loading until a predetermined termination load condition is reached, the termination load condition including the steel box girder entering the yield stage, the occurrence of macroscopic buckling deformation, and reaching the predetermined maximum loading load.

3. The method for identifying hidden damage in steel box girders based on magnetic memory effect as described in claim 1, characterized in that, The original signal also includes: defining magnetic parameters and integral area parameters based on the original signal; The formula for calculating the magnetic parameters is: ， in, For the average value parameter, The total number of testing sites. Let j be the measured value of the normal component of the magnetic signal at the j-th detection point under a specific load. To calculate the arithmetic mean of the magnetic signal values ​​at all detection points; The formula for calculating the integral area parameter is as follows: ， in, For the integral area parameter, and These are the upper and lower limits of the integral. This represents the length increment along the detection line direction. for Along the detection line from arrive Integrate and calculate the area under the curve.

4. The method for identifying hidden damage in steel box girders based on magnetic memory effect as described in claim 3, characterized in that, The magnetic parameters include plotting curves of magnetic parameters as a function of load and analyzing the trend of the curves: The elastic phase curve rises as the load increases; The appearance of a peak point on the yield curve indicates the onset of yielding; The critical stage curve shows a reversal point in the trend at the critical stress state. This reversal point serves as a damage warning sign that the steel box girder has reached the critical stress state. The trend reversal point of the curve refers to the inflection point where the average parameter and the load curve change from rising to falling and from falling to rising.

5. The method for identifying hidden damage in steel box girders based on magnetic memory effect as described in claim 4, characterized in that, The curve trend reversal point also includes: defining key magnetic parameters based on the location of local extreme points; The key magnetic parameters are calculated based on the changes in the normal component of the magnetic memory signal, including: Parameter: The area enclosed by the detection line and the coordinate axis With the length of the detection line The ratio; Parameter: The average value of the absolute value of the magnetic signal on the detection line; Parameter: The maximum absolute value of the magnetic gradient on the detection line; Parameter: The average value of the absolute value of the magnetic gradient on the detection line; Establish a quantitative relationship model between key magnetic parameters and stress at local extreme points.

6. The method for identifying hidden damage in steel box girders based on magnetic memory effect as described in claim 1, characterized in that, The method also includes web bending and shear capacity assessment: defining the average absolute value parameter of the magnetic signal in the web region. ， in, The absolute value parameter of the average magnetic signal. The total number of testing sites. Let be the absolute value of the magnetic signal at the i-th detection point; Establishing dimensionless parameters based on the absolute value parameters of the average magnetic signal: ， in, For dimensionless parameters, The parameter is the absolute value of the average magnetic signal in the initial state. This is the absolute value parameter of the average magnetic signal in the web region under the current load; A linear model is constructed by linearly fitting the dimensionless parameters to the load ratio, which is used to evaluate the bending and shear bearing capacity of the web. The formula for calculating the load ratio is: ,in, For the current load, This is the yield load; The correlation coefficient of the linear model A value greater than 0.85 ensures the reliability of the evaluation results.

7. The method for identifying hidden damage in steel box girders based on magnetic memory effect as described in claim 1, characterized in that, The method further includes multimodal data fusion: deploying acoustic emission sensors, infrared thermal imagers, and fiber Bragg grating sensors to synchronously acquire data with a magnetic memory detector; the deployed acoustic emission sensors are used to capture acoustic signals of microcrack propagation, the infrared thermal imagers are used to monitor temperature anomalies caused by stress concentration, and the fiber Bragg grating sensors are used to measure strain distribution under dynamic loads; the multimodal data are time-stamped and synchronized to construct a comprehensive damage identification dataset.

8. The method for identifying hidden damage in steel box girders based on magnetic memory effect as described in claim 7, characterized in that, The fused multimodal data also includes dynamic evolution modeling: analyzing the evolution trend of damage parameters based on the comprehensive damage identification dataset and calculating the damage index. ， in, The damage index is given by time t. This represents the average value of the magnetic parameters. For acoustic emission energy, For the rate of temperature rise, For strain energy density, The corresponding weighting coefficients were determined through calibration using experimental data, and ; A dynamic damage evolution model is established to describe the damage state transition process, including stable, nascent, and extended states; through the evolution law of the state, early warning of critical yield state and prediction of damage extension are achieved.

9. The method for identifying hidden damage in steel box girders based on magnetic memory effect as described in claim 8, characterized in that, The calculation of the damage index also includes uncertainty quantification: defining uncertainty parameters for each modal feature in the multimodal data, including the measurement error range of the magnetic memory signal, the signal-to-noise ratio of the acoustic emission signal, the temperature error band of the thermal imaging data, and the accuracy error of the strain data; and outputting the mean, variance, and confidence interval of the damage index through a Monte Carlo simulation model.

10. The method for identifying hidden damage in steel box girders based on magnetic memory effect as described in claim 9, characterized in that, The method also includes calculating the overall confidence level: ， in, To assess the overall confidence level, For the average uncertainty, As a modal consistency index, The root mean square error between the damage dynamic evolution model prediction results and historical monitoring data is given. The corresponding weighting coefficients are determined through calibration using historical data and satisfy the following conditions: .

Images