Large-span bridge steel box girder fatigue damage local and overall collaborative evaluation method and system

Abstract

The invention relates to a large-span bridge steel box girder fatigue damage local and overall collaborative evaluation method and system. Acquiring strain responses of a reference section and a test section of the steel box girder before and during traffic service based on a dynamic load test, defining and calculating a strain ratio and a rigidity ratio of each monitoring point, and diagnosing local crack positions of the steel box girder by utilizing statistical characteristics of the strain ratios; establishing a mapping relation between the rigidity ratio and local damage such as crack length, expansion rate and tip position, and proposing a fatigue crack safety coefficient to realize collaborative quantitative characterization of local damage and overall safety; and constructing a genetic algorithm to optimize a support vector machine model, and developing a real-time evaluation system which takes the real bridge strain response as input and takes the local crack damage type and the overall safety level of the steel box girder as output. The defect that local damage diagnosis and overall performance evaluation of the steel box girder are mutually separated in a traditional method is overcome, and real-time diagnosis and safety evaluation of fatigue damage of the large-span bridge can be synchronously achieved.

Description

Technical Field

[0001] This invention relates to the field of safety assessment of bridges in service, and in particular to a method and system for the coordinated assessment of local and overall fatigue damage of steel box girders in long-span bridges. Background Technology

[0002] Long-span bridge steel box girders have complex structures and numerous welds. Under long-term, repeated vehicle loads, fatigue cracks are highly susceptible to occur in weld details and stress concentration areas, leading to structural stiffness degradation, reduced load-bearing capacity, and threatening structural safety. Fatigue crack damage in steel box girders exhibits both distinct local characteristics and affects overall load-bearing performance through load transfer. Therefore, accurate diagnosis of local fatigue crack damage and quantification of its impact on the overall safety of steel box girders are urgently needed to provide a basis for assessing the service condition of bridges.

[0003] Fatigue cracks in steel box girders often originate from structural details such as diaphragms, U-ribs, and top plates. Current methods for diagnosing and assessing fatigue damage in steel box girders mainly include image recognition, ultrasonic guided wave detection, dynamic characteristic analysis, and nominal stress methods or fracture mechanics methods. However, the complex welded structure between steel plates in steel box girders and the presence of severe obstruction significantly limit the practical application of image-based crack identification methods. Due to the complex waveguide propagation paths within thin steel plates, ultrasonic guided wave detection methods can only accurately detect crack damage at known locations. Dynamic response analysis-based fatigue crack damage diagnosis methods for steel box girders require high-order modal data, have stringent requirements for sensor placement, and are highly sensitive to environmental noise. Traditional fatigue life prediction methods rely on crack length monitoring, have limited ability to identify early cracks, and struggle to reveal the synergistic relationship between local damage and overall degradation.

[0004] Furthermore, local structural damage to steel box girders can cause stress redistribution and gradually affect overall load-bearing performance, leading to a decrease in structural stiffness and load-bearing capacity. Existing safety assessment methods largely rely on load-bearing capacity calculations, fatigue life predictions, or statistical early warnings based on monitoring data. Local damage diagnosis and overall performance assessment are separated into two independent parts, lacking a unified indicator system and evaluation model. This makes it difficult to accurately reflect the substantial impact of local crack damage on the overall service safety of steel box girders, hindering the scientific and timely nature of operation and maintenance decisions for large-span steel box girders. Therefore, how to consider the synergistic relationship between local damage and overall safety performance, establish a mapping relationship between the actual bridge load strain response and damage characteristic parameters such as fatigue crack type and location of the steel box girder, and support a comprehensive assessment of the safety status of in-service large-span bridges on this basis, has become an urgent problem to be solved in this field. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for the coordinated assessment of local and overall fatigue damage in steel box girders of long-span bridges, effectively solving the aforementioned technical problems.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] In a first aspect, the present invention provides a method for the coordinated assessment of local and overall fatigue damage in steel box girders of long-span bridges, comprising the following steps:

[0008] Step 1: Select a reference section and a test section on the steel box girder, and arrange several strain sensors as monitoring points at corresponding positions on the two selected sections;

[0009] Step 2: Before the steel box girder is put into service, a dynamic load test is carried out on the steel box girder. The strain response of each monitoring point on the two selected sections is collected in real time by sensors. The strain on the two selected sections is aligned according to the response time using the cross-correlation method.

[0010] Step 3: Define the strain ratio as the ratio of the strain of any monitoring point on the test section to the corresponding monitoring point on the reference section at a certain moment; based on the strain response after alignment in Step 2, calculate the strain ratio of any monitoring point on the test section to the corresponding monitoring point on the reference section at different moments, and use data statistics methods to obtain the statistical characteristic value of the strain ratio of any monitoring point before service.

[0011] Step 4: After the steel box girder has been put into service, conduct a bridge dynamic load test on the steel box girder at the current service time, and obtain the statistical characteristic value of the strain ratio at any monitoring point at the current service time according to the methods in Steps 2 and 3.

[0012] Step 5: Define the stiffness ratio as the ratio of the strain ratio at any monitoring point on the test section of the steel box girder at the current service time to the strain ratio before service. Calculate the stiffness ratio of any monitoring point on the test section at the current service time based on the statistical characteristic values ​​of the two types of strain ratios obtained in Steps 3-4. Based on the stiffness ratio distribution on the same section, take the location with the largest change in absolute stiffness ratio as the crack damage location, and determine the crack damage type based on the damage location.

[0013] Step 6: Perform differential processing on the stiffness ratio obtained in Step 5. Based on the differential stiffness ratio data, use a genetic algorithm to optimize the support vector machine (GA-SVM) model to obtain different fatigue crack damage types and corresponding fatigue crack safety levels of the steel box girder.

[0014] It should be noted that in step 1, the section with the largest load response and the most likely to cause fatigue damage among all sections of the steel box girder is selected as the test section. The test section is usually a section within the range of 1 / 3 to 2 / 3 of the span of the steel box girder. The section with a relatively large load response but not prone to fatigue damage among all sections of the steel box girder is selected as the reference section. The reference section is usually any section within the range of 1 / 5 to 1 / 3 or 2 / 3 to 4 / 5 of the span of the steel box girder.

[0015] It should be noted that in step 1, the fatigue crack damage of the steel box girder includes three types: Type I damage is transverse cracking at the weld between the diaphragm and the U-rib and at the opening of the diaphragm; Type II damage is transverse cracking at the weld between the diaphragm and the top plate; and Type III damage is longitudinal cracking at the weld between the U-rib and the top plate. Strain monitoring points are set at the bottom of the U-rib in the longitudinal direction along the bridge and at the opening of the diaphragm in the vertical direction along the bridge.

[0016] It should be noted that in steps 3-4, the strain ratio is obtained according to equations (3.1) and (4.1):

[0017] (3.1)

[0018] (4.1)

[0019] In the formula, Reference section The measured strain value at monitoring point i at a certain moment; Test section of steel box girder before service The measured strain value at monitoring point i at a certain moment; Test section for the current service time T of the steel box girder The measured strain value at monitoring point i at a certain moment; The strain ratio of monitoring point i on the test section of the steel box girder before service relative to monitoring point i on the reference section; The strain ratio of monitoring point i on the test section relative to monitoring point i on the reference section is given by the current service time T of the steel box girder. This is the distance between a certain section on the steel box girder and the end section.

[0020] It should be noted that in step 3, the data statistical method adopts the strain ratio probability density function; the statistical feature value adopts one of the mode, maximum value, and mean value; preferably the mode.

[0021] It should be noted that in step 5, the stiffness ratio is obtained according to equation (5.1):

[0022] (5.1)

[0023] In the formula, The stiffness ratio of monitoring point i on the test section at the current service time T of the steel box girder (relative to monitoring point i on the test section before service).

[0024] It should be noted that in step 6, the differentiation process is as follows: first, the original stiffness ratio data is subjected to a nonlinear transformation to amplify the differences in data features; then, the stiffness ratio data after the nonlinear transformation is standardized to make the data have the same scale; finally, the standardized stiffness ratio data is extended by a polynomial to increase the complexity of the data and obtain differentiated stiffness ratio data; the polynomial is preferably a binomial.

[0025] It should be noted that step 6, which uses a genetic algorithm to optimize the support vector machine model to obtain the fatigue crack damage type and fatigue crack safety level of the steel box girder, includes the following steps:

[0026] 1) Based on the stiffness ratio data under the current service time, substitute it into the mapping model of stiffness ratio with crack length, crack propagation rate and crack tip position to obtain the crack length, crack propagation rate and crack tip position under the current service time.

[0027] The mapping model between the stiffness ratio of any monitoring point in the test section and the crack length, crack propagation rate, and crack tip position is obtained by fitting based on indoor crack monitoring tests or field crack monitoring tests, and is established according to equation (6.1):

[0028] (6.1)

[0029] In the formula, , , These are the length, propagation rate, and tip position of crack j under the current service time T; , , These represent the mapping relationships between stiffness ratio and crack length, crack propagation rate, and crack tip location, respectively, obtained by fitting based on measured data; Φ is a comprehensive function, which can be implemented using response surface methodology, Gaussian process, Kriging model, and other methods to achieve multi-parameter model spatial regression fitting.

[0030] 2) Calculate the fatigue crack safety factor for the current service time based on the crack length, crack propagation rate, and crack tip position.

[0031] The fatigue crack safety factor is obtained according to formula (6.2):

[0032] (6.2)

[0033] In the formula, The fatigue crack safety factor; , , Current service time Scores for the length, propagation rate, and tip position of the lower crack j

[0034] Among them, The score is determined as follows:

[0035] is the score corresponding to the fatigue crack length. When the crack length is greater than 0 and less than 100 mm, that is when = 1; when the crack length is greater than or equal to 100 mm and less than 150 mm, that is when = 2; when the crack length is greater than or equal to 150 mm, that is when = 3;

[0036] Among them, The score is determined as follows:

[0037] is the score corresponding to the fatigue crack propagation rate. When the crack does not propagate, that is when = 1; when the crack propagation rate is greater than 0 and less than 5 mm / month, that is when = 2; when the crack propagation rate is greater than 5 mm / month, that is when = 3;

[0038] Among them,​​​​​​​​​​​​​​​​​​​​​​​​​​​​Secondly, the present invention provides a system for the coordinated assessment of local and overall fatigue damage of steel box girders for long-span bridges, used to perform the aforementioned method for the coordinated assessment of local and overall fatigue damage of steel box girders for long-span bridges, including:

[0043] The data acquisition module is used to: collect the strain response of each monitoring point on the reference section and test section of the steel box girder before and during the current service time in real time;

[0044] The data processing module is used to: calculate the stiffness ratio of each monitoring point on the test section at the current service time based on the strain response; and use the GA-SVM algorithm model to obtain the fatigue crack type and fatigue crack safety level of the steel box girder based on the differentially processed stiffness ratio data.

[0045] The data storage module is used to store monitoring data, data during data processing, and data on fatigue crack types and fatigue crack safety levels of steel box girders.

[0046] The data output module is used to output the fatigue crack type of the steel box girder, as well as the corresponding fatigue crack safety level and countermeasures.

[0047] It should be noted that, based on the fatigue crack safety level of the steel box girder, the countermeasures are determined according to the following method:

[0048] When the safety level is A, the response measure is to conduct regular observation of the cracks at least once every six months;

[0049] When the safety level is B, the response is to closely observe the cracks at least once a month.

[0050] When the safety level is C, the response is to promptly repair and reinforce the cracks.

[0051] Compared with the prior art, the beneficial effects of the present invention are reflected in:

[0052] (1) By integrating the real-time response of strain measurement points deployed on the actual bridge, and combining cross-correlation function alignment and strain ratio probability density analysis, a multi-dimensional damage characteristic parameter system is formed to achieve sensitive capture and location identification of different types of crack damage conditions, thereby providing scientific basis and real-time decision support for intelligent diagnosis and safety level assessment of fatigue cracking of steel box girders.

[0053] (2) A crack damage local stiffness degradation model based on the strain ratio and its derived stiffness ratio distribution at each strain monitoring point of the steel box girder is proposed. This model quantifies the influence of local crack damage on the overall stiffness of the steel box girder and breaks through the limitation of the traditional separation between local damage diagnosis and overall safety assessment.

[0054] (3) By using the fatigue crack importance coefficient, the multi-feature factors of local damage such as crack length, propagation rate and tip position are comprehensively quantified, realizing the quantitative correlation and synergy between local damage and overall safety, breaking through the subjective limitation of relying on experience judgment, and supplemented by the data feature difference enhancement method and GA-SVM intelligent algorithm to improve the accuracy of local damage diagnosis and overall safety assessment. Attached Figure Description

[0055] Figure 1 This is a flowchart of the method of the present invention;

[0056] Figure 2 This is a diagram showing the structural dimensions of the steel box girder and the lane layout.

[0057] Figure 3 Cross-sectional view of the monitoring point layout for the steel box girder;

[0058] Figure 4 This is a schematic diagram of three typical fatigue damage types in steel box girders;

[0059] Figure 5 Figure 1 shows the strain time history curves of the strain monitoring points of the U-rib of the steel box girder; Figure 2 shows the Span2 section and Figure 3 shows the Span3 section.

[0060] Figure 6 Figure 1 shows the strain time history curves of the strain monitoring points of the transverse diaphragm of the steel box girder; Figure 2 shows the DP2 section and Figure 3 shows the DP3 section.

[0061] Figure 7 Figure 1 shows the alignment curves of strain monitoring points for the U-ribs and transverse diaphragms of the steel box girder; Figure 2 shows the strain measurement points of the U-ribs along the bridge direction, and Figure 3 shows the strain measurement points of the transverse diaphragms along the bridge direction.

[0062] Figure 8 This is a schematic diagram showing the relationship between Type I crack damage and the location of monitoring points;

[0063] Figure 9 This is a schematic diagram of the longitudinal strain ratio probability density distribution of the U-rib under type I crack damage in a steel box girder.

[0064] Figure 10 This is a schematic diagram of the vertical strain ratio probability density distribution of the transverse diaphragm under Type I crack damage in a steel box girder;

[0065] Figure 11 A schematic diagram showing the relationship between type II and type III fatigue crack damage and the location of monitoring points in a steel box girder;

[0066] Figure 12 This is a schematic diagram of the longitudinal strain ratio probability density distribution of the U-rib under type II crack damage in a steel box girder.

[0067] Figure 13This is a schematic diagram of the vertical strain ratio probability density distribution of the transverse diaphragm under type II crack damage in a steel box girder.

[0068] Figure 14 This is a schematic diagram of the longitudinal strain ratio probability density distribution of the U-rib under type III crack damage in a steel box girder.

[0069] Figure 15 This is a schematic diagram of the vertical strain ratio probability density distribution of the transverse diaphragm under type III crack damage in a steel box girder.

[0070] Figure 16 This is a schematic diagram showing the stiffness ratio distribution of the U-rib section at mid-span of the Span 3 section under different crack lengths in a steel box girder with type III crack damage.

[0071] Figure 17 A schematic diagram of the technical framework for optimizing the Support Vector Machine (GA-SVM) algorithm model using genetic algorithms;

[0072] Figure 18 A schematic diagram of the framework for assessing the fatigue cracking safety performance of steel box girders;

[0073] Figure 19 Figure 1 shows the results of fatigue crack type diagnosis and damage safety level assessment. Figure 2 shows the fatigue crack type diagnosis results without differentiation treatment, Figure 3 shows the fatigue crack damage safety level assessment results without differentiation treatment, Figure 4 shows the fatigue crack type diagnosis results with differentiation treatment, and Figure 5 shows the fatigue crack damage safety level assessment results with differentiation treatment. Detailed Implementation

[0074] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0075] This invention provides a method and system for the coordinated assessment of local and overall fatigue damage in long-span steel box girders. Driven by the need for safety performance assessment of in-service long-span steel box girders, it considers the mechanism of overall service safety degradation induced by local damage. Based on strain sensor monitoring point data processing and optimal selection of damage characteristic parameters (strain ratio), it integrates a genetic algorithm-optimized support vector machine (GA-SVM) algorithm model to drive a steel box girder fatigue cracking safety performance assessment method, achieving online monitoring and safety assessment of fatigue cracking damage in in-service steel box girders. The method and system architecture mainly include three layers: equipment layer, algorithm layer, and application layer. The equipment layer mainly includes standard vehicles for bridge load testing, various types of sensors, and corresponding data acquisition equipment. The algorithm layer mainly includes sensor measured data processing methods, strain ratio and stiffness ratio and stiffness degradation modeling, GA-SVM algorithm model construction and testing training, and other theoretical methods. The application layer mainly includes a data acquisition module, a data processing module, a data storage module, and a data output module.

[0076] Implementation method one.

[0077] Please see Figure 1 This embodiment provides a method for the coordinated assessment of local and overall fatigue damage in steel box girders of long-span bridges, including the following steps:

[0078] Step (1): Select the reference section and test section on the steel box girder, and arrange several strain sensors as monitoring points at corresponding positions on the two selected sections;

[0079] When conducting fatigue crack damage diagnosis and safety performance assessment of steel box girders, it is first necessary to determine the reference section and the test section. In actual engineering, the section with the largest load response and the highest likelihood of fatigue damage among all sections of the steel box girder should be selected as the test section. The test section is usually located within 1 / 3 to 2 / 3 of the span of the steel box girder. The reference section is selected from sections other than the test section that have a large load response and no structural defects (not prone to fatigue damage). The reference section is usually located within 1 / 5 to 1 / 3 or 2 / 3 to 4 / 5 of the span of the steel box girder, preferably at 1 / 4 or 3 / 4. The strain response at the reference section and the test section of the steel box girder is the strain response data generated by vehicle wheel loads. Multiple test sections can be selected; it is best to choose the section containing the transverse diaphragm or the mid-span section of the U-rib between two transverse diaphragms as the test section.

[0080] The main locations for strain monitoring points on the two cross sections of the steel box girder are: near the weld between the diaphragm and the U-rib and its opening, and at the bottom of the U-rib mid-span, where stress is significant or prone to cracking. The locations of the monitoring points can also be optimized and adjusted according to actual engineering needs. Typical fatigue damage defects in steel box girders include transverse cracking at the weld between the diaphragm and the U-rib and its opening, longitudinal cracking at the weld between the U-rib and the top plate, and transverse cracking at the weld between the diaphragm and the top plate. We can assume that transverse cracking at the weld between the diaphragm and the U-rib and its opening is Type I damage, transverse cracking at the weld between the diaphragm and the top plate is Type II damage, and longitudinal cracking at the weld between the U-rib and the top plate is Type III damage.

[0081] Strain monitoring points are installed along the bridge direction at the bottom of the U-rib span and vertically at the opening of the transverse diaphragm. "Vertical" and "horizontal" refer to the orientation of the strain sensors; for example, "vertical" means the sensors are attached vertically to the surface of the transverse diaphragm. As a compression-bending member, the vertical stress characteristics of the transverse diaphragm can be effectively obtained through this arrangement to obtain its strain response. The number of monitoring points is usually consistent with the number of U-ribs under the lane; for example, if there are six U-ribs under a lane, six monitoring points are installed. It should be noted that these monitoring points are not specifically designed for a particular type of crack damage, but rather utilize the strain monitoring points arranged on the U-ribs and transverse diaphragms to reflect the occurrence and development characteristics of different types of crack damage.

[0082] Step (2): Before the steel box girder is put into service, a dynamic load test is carried out on the steel box girder. The strain response of each monitoring point on the two selected sections is collected in real time by the sensor. The strain on the two selected sections is aligned according to the response time by the cross correlation method.

[0083] Dynamic load testing of bridges is a standard technique in this field. It involves marking lanes on the bridge and deploying sensors at specified locations. Then, a standard vehicle meeting the test axle load is selected and driven from a predetermined starting position to a predetermined ending position at a specified speed, during which data is collected. Finally, depending on the data collected, additional test conditions are added, or the data is directly processed. The axle load of the standard vehicle is generally calculated based on the test requirements and the bridge span layout; different axle loads are selected for different test purposes.

[0084] When the same vehicle load passes through several sections of the steel box girder in sequence along the driving direction, there will be a time delay in the wheel load strain time history generated at the monitoring points corresponding to the reference section and the test section. The cross-correlation method can be used to align the strain on the two selected sections according to the response time, that is, to align the strain monitoring data using the cross-correlation function method, so as to ensure the accuracy of the strain ratio calculation at each measuring point.

[0085] The cross-correlation function is a function that measures the change in similarity between two signals over time. For two discrete signals W[n] and U[n], their cross-correlation function Rn... xy [l] is defined as:

[0086] (1)

[0087] In the formula, l is the delay parameter, representing the number of samples by which signal U[n] is shifted relative to signal W[n]. When two discrete signals W[n] and U[n] are aligned in time, their cross-correlation function will reach its maximum value, which corresponds to the phase difference between the two sets of signals.

[0088] Step (3): Define the ratio of the strain of any monitoring point on the test section to the corresponding monitoring point on the reference section at a certain moment as the strain ratio; according to the strain response after alignment in step (2), calculate the strain ratio of any monitoring point on the test section to the corresponding monitoring point on the reference section at different times, and use data statistics methods to obtain the statistical characteristic value of the strain ratio of any monitoring point before service.

[0089] Step (4): After the steel box girder is put into service, a bridge dynamic load test is carried out on the steel box girder at the current service time, and the statistical characteristic value of the strain ratio of any monitoring point at the current service time is obtained according to the methods of steps (2) and (3).

[0090] The mid-span region of long-span steel box girders exhibits a significant positive bending moment response and is a high-risk area for fatigue cracking. Therefore, deploying sensors at the mid-span test section to acquire strain data can effectively reflect local fatigue damage characteristics. In contrast, sections near the supports have complex stress states, and their strain is easily affected by boundary constraints and local force transmission, making them unsuitable as reference sections. Therefore, selecting a section with a relatively simple stress state and a large load response within the 1 / 5-1 / 3 span or 2 / 3-4 / 5 span range of the steel box girder as the reference section for sensor deployment can better reflect the strain characteristics under normal service conditions. Comparing the strain responses obtained at each measuring point on the test section with those obtained at the corresponding measuring points on the reference section, the resulting strain ratio distribution reflects the relative relationship between the local and overall strain characteristics of the steel box girder. Since the stress characteristics of the reference section change little after local damage occurs at other sections, the strain can be considered constant. Therefore, the stiffness ratio derived from the strain ratio can accurately reflect the impact of local damage on the overall stiffness, overcoming the subjective limitations of relying on empirical judgment.

[0091] Furthermore, the principle of using strain ratio distribution for local crack damage diagnosis of steel box girders is as follows: When fatigue cracks appear in a steel box girder, the local stiffness of the crack-damaged area will significantly decrease. Under vehicle loads, the strain response of the damaged area will differ from that of the undamaged area. By placing strain sensors on key sections of the steel box girder, strain data of the structure under different working conditions is collected, and the strain ratio between adjacent measuring points is calculated based on this data. Because fatigue cracks alter the stress transmission path and local stiffness distribution of the steel box girder, the strain ratio near the damaged area will change significantly, thereby allowing for the diagnosis of the location of local crack damage in the steel box girder.

[0092] The damage characteristic parameters of the steel box girder are preferably unaffected by fluctuations in external temperature and load amplitude, and can sensitively capture the strain ratio and its derived stiffness ratio of the local damage state of the structure. Based on the bridge influence line theory, the expression for calculating the strain ratio of the test section relative to the reference section is:

[0093] (2)

[0094] In the formula, Pre-service test section The strain ratio relative to the reference section; and These are the reference sections of the steel box girder before service. and test cross section Measured strain value at the location; and To be laid out at the reference section of the steel box girder and test cross section The sensitivity coefficient of the strain sensor at the location; L is the calculated span of the steel box girder. Because L, , , , Given the values ​​of the parameters, and assuming the bridge is undamaged, under the same load, the test sections of the bridge... The strain ratio will not change, that is:

[0095] (3)

[0096] In the formula, k is the test cross section. The sequence number, k=1, 2, ..., n.

[0097] By monitoring each test section Changes in strain ratio can be used to diagnose damage to bridges. Assuming a steel box girder test section... Damage occurred at the location, generally considered to be at the reference section. The location is at the test section The strain ratio before and after damage remains unchanged, that is... Test section Relative to the reference section The strain ratio can be written as:

[0098] (4)

[0099] In the formula, Test section for current service time T The strain ratio relative to the reference section; and These are the reference sections of the steel box girder before service. and test cross section The measured strain value at the location.

[0100] At the test section Reference section Multiple monitoring points are set up on the steel box girder, and the positions of the monitoring points correspond to those on the cross-section of the steel box girder. Therefore, for the test cross-section... The strain ratio at monitoring point i is obtained according to equation (5):

[0101] , (5)

[0102] In the formula, Reference section The measured strain value at monitoring point i at a certain moment; Test section of steel box girder before service The measured strain value at monitoring point i at a certain moment; Test section for the current service time T of the steel box girder The measured strain value at monitoring point i at a certain moment; The strain ratio of monitoring point i on the test section of the steel box girder before service relative to monitoring point i on the reference section; The strain ratio of monitoring point i on the test section relative to monitoring point i on the reference section is given by the current service time T of the steel box girder. This is the distance between a certain section on the steel box girder and the end section.

[0103] In bridge dynamic load tests, the strain response collected by strain sensors is time-history data, i.e., a curve of strain changing over time. It is difficult to directly apply the strain to the reference section. Strain response and test cross section at a certain monitoring point By comparing the strain responses at the corresponding monitoring points, it is possible to select several strain values ​​at the same time on these two aligned strain responses for comparison. Based on the corresponding multiple strain values, multiple strain ratios can be obtained.

[0104] Then, statistical methods are used to obtain the statistical characteristic values ​​of the strain ratio before service at each monitoring point. The statistical method employs a probability density function, establishing a strain ratio probability density function based on the aforementioned multiple strain ratios, thereby determining the statistical characteristic values ​​of the strain ratio. The statistical characteristic values ​​can be the mode, maximum, or mean; this invention preferably uses the mode, which is used as the basis for subsequent stiffness ratio calculations.

[0105] The method for obtaining the statistical characteristic value of the strain ratio at any monitoring point during the current service time of the steel box girder is the same as that in steps (2) and (3). For newly built steel box girder bridges, after the first dynamic load test of the bridge is carried out before service, the test is repeated every 1 to 2 years in the first few years, and after 5 years, the test is repeated every 0.5 to 1 year.

[0106] Step (5): Define the stiffness ratio as the ratio of the strain ratio of any monitoring point on the test section of the steel box girder at the current service time to the strain ratio before service. Calculate the stiffness ratio of any monitoring point on the test section at the current service time based on the two types of statistical characteristic values ​​obtained in steps (3) and (4). Based on the stiffness ratio distribution on the same section, take the location with the largest change in absolute stiffness ratio as the crack damage location, and determine the crack damage type based on the damage location.

[0107] Under the same load conditions, the strain ratio of the same cross section before service and the current service time reflects the relative change in structural stiffness, which is described by the stiffness ratio (β).

[0108] (6)

[0109] In the formula, Test section for current service time T Compared to the pre-service test section stiffness ratio; , Test cross sections Stiffness before and during the current service time T.

[0110] Therefore, the stiffness ratio of a certain monitoring point on the test section is obtained according to equation (7):

[0111] (7)

[0112] In the formula, Test section for the current service time T of the steel box girder Upper monitoring point i relative to the pre-service test section The stiffness ratio of the upper monitoring point i.

[0113] Before the steel box girder was damaged, It remains unchanged, with a value of 1. When the steel box girder is damaged, stress redistribution will occur at the test section. It will change; some values ​​will be less than 1, and some values ​​will be greater than 1. Stiffness ratio after damage. The more pronounced the change, the greater the probability of fatigue damage occurring near the corresponding monitoring point. Based on this, the location of the monitoring point with the largest change in the absolute value of the stiffness ratio can be preliminarily determined as the location of or near the fatigue crack, and this information should be recorded.

[0114] Step (6): Differentiate the stiffness ratio obtained in step (5). Based on the differentiated stiffness ratio data, use the genetic algorithm to optimize the support vector machine (GA-SVM) model to obtain different fatigue crack damage types and corresponding fatigue crack safety levels of the steel box girder.

[0115] The stiffness ratio distribution data of steel box girders after localized cracking damage exhibits relatively small differences. The high similarity between data features poses a challenge to the classification task of the GA-SVM algorithm model, necessitating differential processing of the original stiffness ratio data features. This differential processing includes three steps: first, performing a nonlinear transformation on the original stiffness ratio data to amplify the differences in data features; second, standardizing the data after the nonlinear transformation to ensure uniform scale; and finally, expanding the standardized data using a polynomial to increase data complexity, resulting in differentiated stiffness ratio data. The preferred polynomial is a binomial.

[0116] The process of using a genetic algorithm to optimize a support vector machine (GA-SVM) model to obtain the fatigue crack damage type and fatigue crack safety level of a steel box girder includes the following steps:

[0117] 1) Based on the stiffness ratio data under the current service time, substitute it into the mapping model of stiffness ratio with crack length, crack propagation rate and crack tip position to obtain the crack length, crack propagation rate and crack tip position under the current service time.

[0118] The mapping model between the stiffness ratio of any monitoring point and the crack length, crack propagation rate, and crack tip position is obtained by fitting the model based on indoor crack monitoring tests or field crack monitoring tests, and is established according to equation (8):

[0119] (8)

[0120] In the formula, , , These are the length, propagation rate, and tip position of crack j under the current service time T; , , These represent the mapping relationships between stiffness ratio and crack length, crack propagation rate, and crack tip location, respectively, obtained by fitting based on measured data; Φ is a comprehensive function, which can be implemented using response surface methodology, Gaussian process, Kriging model, and other methods to achieve multi-parameter model spatial regression fitting.

[0121] The fitting function can be selected based on the changing trend of the measured data, and the simplest function with the highest fitting accuracy should be chosen as much as possible.

[0122] Indoor crack monitoring test: Steel box girders are usually made of low alloy steel plates, such as Q345, Q355, Q550 series steel, etc., and are manufactured by full welding. Indoor tests can be conducted on welded steel plates with the same material properties. Different crack lengths and different crack tip positions are set on the welded steel plates. Then, several strain gauges are arranged in the area near the crack, and fatigue load tests are carried out. By changing the fatigue load amplitude or frequency, different crack propagation rates can be simulated. Finally, by collecting strain data in real time, strain ratio or stiffness ratio is formed, and then a corresponding mapping model can be established based on this.

[0123] Field Crack Monitoring Test: For a steel box girder bridge in service, when fatigue cracking was found at a specific detail during routine inspection, several strain gauges were placed in this area. Then, over a period of time, the crack length and crack tip location were continuously monitored at a certain observation frequency. It should be noted that this bridge was not the subject of a fatigue damage safety assessment, but rather other in-service bridges with similar structural forms. The primary purpose was to obtain real-world measured data required for the mapping model. The crack propagation rate can be calculated based on the crack length at different times. Then, the strain response data at each detection frequency is recorded. From this, strain ratios or stiffness ratios can be derived by real-time acquisition of strain data, and a corresponding mapping model can be established based on these ratios.

[0124] 2) Calculate the fatigue crack safety factor for the current service time based on the crack length, crack propagation rate, and crack tip position.

[0125] The fatigue crack safety factor (λ) comprehensively considers multiple key factors such as crack length, propagation rate, and crack tip location. It quantifies the hazard level of fatigue cracks through specific numerical values, directly reflecting the impact of localized fatigue cracks on the overall structural safety, thus avoiding the shortcomings of assessments based solely on subjective experience. The formula for calculating the fatigue crack safety factor is:

[0126] (9)

[0127] In the formula, The fatigue crack safety factor; , , Current service time The scores for the length, propagation rate, and tip position of the lower crack j.

[0128] in, The score is determined as follows:

[0129] This represents the score corresponding to the fatigue crack length; when the crack length is greater than 0 and less than 100 mm, it is considered a fatigue crack. hour, =1; when the crack length is greater than or equal to 100mm and less than 150mm, i.e. hour, =2; when the crack length is greater than or equal to 150mm, i.e. hour, =3;

[0130] in, The score is determined as follows:

[0131] is the score corresponding to the fatigue crack growth rate. When the crack does not grow, i.e., when, = 1; when the crack growth rate is greater than 0 and less than 5 mm / month, i.e., when, = 2; when the crack growth rate is greater than 5 mm / month, i.e., when, = 3;

[0132] Among them, The score of is determined according to the following method:

[0133] is the score corresponding to the position of the fatigue crack tip. When the crack tip is located at the weld toe or weld root, i.e., = H or 0, = 1; when the crack tip is located within the weld height, i.e., 0 < < H, = 2; when the crack tip is located in the base metal, i.e., < 0 or > H, = 3; H is the weld height;

[0134] 3) According to the fatigue crack safety factor under the current service time, output the fatigue crack safety level;

[0135] According to the magnitude of the fatigue crack safety factor (λ), the fatigue crack damage can be divided into different safety levels: when 0.3 ≤ λ < 0.5, the safety level is A level; when 0.5 ≤ λ < 0.7, the safety level is B level; when 0.7 ≤ λ ≤ 1.0, the safety level is C level.

[0136] By performing tests and evaluations according to steps 4 to 6, the types of fatigue crack damage that may occur in the steel box girder at any service time and their corresponding fatigue crack safety levels can be obtained.

[0137] When there are multiple test sections arranged on a steel box girder, multiple fatigue crack safety factors (λ) will be obtained. Select the safety level corresponding to the maximum fatigue crack safety factor (λ max ) as the evaluation result of the current safety state or technical condition of the bridge, and the maintenance plan for each test section is implemented according to the corresponding countermeasures of the safety level of this section.

[0138] The technical approach of this invention is as follows: By statistically analyzing the strain ratio distribution at each monitoring point in a steel box girder, a strain ratio probability density function is constructed for each measuring point under different types of crack damage. Statistical parameters such as the mode of the strain ratio at each measuring point under different types of crack damage conditions are obtained. The stiffness ratio is calculated based on the strain ratio statistical parameters before and during the current service period. Based on the differentiated stiffness ratio data, a support vector machine (SVM) model is optimized using a genetic algorithm (GA) to obtain the optimal solution for the penalty factor C and the kernel function parameter g. The fatigue crack safety level and fatigue crack type are then determined. Using the strain ratio and stiffness ratio distribution information at each monitoring point in the steel box girder, accurate diagnosis of fatigue crack damage type and quantitative assessment of structural safety level of the steel box girder are achieved.

[0139] Typically, a bridge has multiple lanes, let's say the number of lanes is m. When all m lanes are occupied by vehicles, the strain response of a monitoring point placed under one lane will be affected by the vehicle loads from the other m-1 lanes. Therefore, the stiffness ratio matrix is ​​actually a multi-dimensional matrix. The stiffness ratio matrix after bridge structure damage is then:

[0140] (10)

[0141] In the formula, C represents all information of the bridge dynamic load test vehicle before service; C' represents all information of the bridge dynamic load test vehicle at a certain service time; the test vehicle information includes vehicle model, speed, wheelbase, and axle load.

[0142] Furthermore, local damage conditions of steel box girders include fatigue cracking, bolt detachment, anchor bolt fracture, and failure of welded and other connecting components. The overall structural performance parameters of the steel box girder include physical quantities or mechanical performance indicators such as overall structural stiffness and deformation, low-order structural modes, structural fatigue life, residual structural strength, structural reliability, and stability coefficient. This invention uses fatigue cracking as a local damage condition and overall structural stiffness and strain as performance indicators, providing a fatigue damage assessment method for steel box girders that coordinates local damage with overall safety. For other local damage conditions, corresponding performance indicators can be selected, and corresponding fatigue damage assessment methods can be established according to the technical approach provided by this invention.

[0143] For localized damage conditions caused by bolt detachment, this invention can select low-order structural modes and connection stiffness degradation as performance indicators. Bolt detachment mainly leads to changes in local connection stiffness and overall vibration characteristics. Its damage effect is more sensitive to frequency changes and modal coupling, unlike fatigue cracking, which mainly causes local strain anomalies and overall stiffness reduction through cross-sectional weakening.

[0144] For anchor bolt fracture cases, the structural residual strength and overall stability coefficient can be selected as performance indicators. Anchor bolt fracture leads to a decrease in load-bearing capacity reserve and a reduction in the critical load for instability. The key is to assess the continuity of the load-bearing capacity transfer path and the overall stability performance, which differs from fatigue cracking, which focuses more on local crack propagation and stress concentration.

[0145] In cases of fracture failure of welded or other connected components, the present invention can use structural reliability and fatigue life indices as performance parameters. Weld fracture is more sudden and brittle, often manifested as a sharp drop in load-bearing capacity and a drastic shortening of fatigue life. Its diagnostic characteristics tend to focus on transient response and reliability loss, which differs from the progressive evolutionary characteristics of fatigue crack propagation. Specific Implementation Example 1:

[0147] Combination Figures 1 to 15 This embodiment takes a steel bridge on a highway ramp as an example to provide a method for assessing fatigue damage of steel box girders that combines local and overall factors. The specific content is as follows:

[0148] A highway ramp bridge consists of a 20m long and 10.5m wide steel box girder, with two lanes in each direction, each lane being 3.75m wide. Based on the lane division of the bridge deck, the U-ribs under each lane are numbered "U1" to "U6". The diaphragms of the steel box girder are spaced 4m apart, and are numbered "DP1" to "DP6" along the direction of traffic. Figure 2 As shown, a refined finite element numerical simulation model of the actual bridge, with shell elements of the same size, was established using the finite element software ABAQUS. The elasticity of the steel was 210 GPa, Poisson's ratio was 0.3, and density was 7850 kg / m³. 3 The vehicle load was calculated using the Type III fatigue load model in the "Design Code for Highway Steel Structure Bridges" (JTG D64-2015). The vehicle speed was 1 m / s, and the calculation time was set to 30 s. The Dload subroutine, written in Fortran, was integrated into ABAQUS to simulate the vehicle load movement.

[0149] It is important to note that each dynamic load test of a bridge requires traffic interruption and the selection of standard vehicles that meet the test axle load requirements. However, traffic interruption after the steel box girder enters service has a significant impact on normal social production and daily life. Therefore, in practical engineering, a combination of on-site monitoring and numerical simulation can be used. However, when using numerical simulation, the numerical model needs to be calibrated based on on-site monitoring data to ensure the accuracy of the monitoring results obtained from the numerical model. In this specific embodiment, for the convenience of case illustration and verification, the numerical simulation results with reasonable assumptions and parameter settings are used as the basis for case analysis to prove the accuracy of the safety assessment method described in this invention.

[0150] Strain monitoring points were set at the bottom of the U-rib span along the longitudinal direction and at the opening of the transverse diaphragm along the vertical direction, such as... Figure 3 As shown. The naming rule for strain measurement points on the diaphragm is "DP'X'-1 to 6", such as the strain measurement point near the third U-rib in the first diaphragm DP1 of the steel box girder being "DP1-3"; the area between the two diaphragms of the steel box girder is defined as a Span (inner span), and the naming rule for strain measurement points at the bottom of the U-rib is "Span 'X'-U-1 to 6", such as the strain measurement point at the bottom of the second U-rib in the first Span of the steel box girder being "Span 1-U-2". In this embodiment, the reference section of the diaphragm is section DP2, and the reference section of the U-rib is the mid-span position of Span 2. The test section of the diaphragm is section DP3, and the test section of the U-rib is the mid-span position of Span 3. The strain measurement points of the test sections and the reference sections correspond one-to-one in cross-sectional position.

[0151] This embodiment sets up three fatigue cracking damage conditions for steel box girders, and their specific locations are as follows: Figure 4 As shown in the diagram. Type I damage is transverse cracking of the diaphragm; Type II damage is transverse cracking at the weld between the diaphragm and the top plate; and Type III damage is longitudinal cracking at the weld between the U-ribs and the top plate. The crack lengths for Type I damage are set to 60mm, 120mm, and half the distance between the two U-ribs (approximately 225mm). The crack lengths for Type II damage are 60mm, 120mm, and penetrating between the two U-ribs (approximately 270mm). The crack lengths for Type III damage are set to 60mm, 120mm, and 200mm. Assuming a crack width of 2mm and a crack depth penetrating the steel plate thickness, crack propagation is considered by setting different crack length conditions, without considering structural nonlinearity.

[0152] Figure 5 and Figure 6 These are the strain-time history curves generated at monitoring points when the vehicle load passes through several cross sections sequentially along the driving direction. Under the same vehicle load, Figure 5 and Figure 6 The wheel-load strain responses exhibit time lag and similar morphologies. A cross-correlation method is used to align the strain monitoring data. Figure 7 The alignment curves are shown for the strain measurement points of the U-ribs and the strain monitoring points of the transverse diaphragms of the steel box girder under vehicle load.

[0153] After aligning and normalizing the strain data of the vehicle load passing through the steel box girder, the mid-span bottom section of the U-rib in the Span 2 section and the DP2 transverse diaphragm section were used as the reference sections for calculation. Figure 2 The probability density model of wheel load strain ratio at each measuring point is used as a basis for further analysis. Figure 4 The damage of the steel box girder under three fatigue crack conditions is diagnosed. Figure 4Let's take type I injury as an example to illustrate. Figure 8 This is a schematic diagram showing the transverse crack reaching 60mm (Type I crack damage) at the opening between the third U-rib and the transverse diaphragm in section DP3. Figure 9-10 This represents the wheel load strain ratio distribution at each monitoring point of the reference section and the test section. Figure 8 and Figure 9-10 It can be seen that the strain response of the measuring points located in the vertical bridge direction of the transverse diaphragm and the longitudinal bridge direction at the bottom of the U-rib under vehicle load is more sensitive to local crack damage. For example, the wheel load strain ratio distribution at the Span 3-U-3 measuring point at the bottom of the U-rib has changed to a certain extent compared with the reference section Span 2-U-3. The wheel load strain ratio distribution at the DP3-3 measuring point of the transverse diaphragm has changed significantly compared with the reference section DP2-3. The strain ratio distribution of the other measuring points has not changed significantly. Transverse cracks in the transverse diaphragm can be identified by the strain ratio of the transverse diaphragm monitoring points near the crack damage and the strain ratio of the measuring points at the bottom of the nearby U-rib. The above results show that the strain ratio method can infer fatigue detail cracking near the DP3-3 measuring point, which is consistent with the damage state of the actual structure. By integrating the strain ratio probability density distribution of each monitoring point, the mode values ​​of each measuring point under normal and damaged conditions are obtained, as shown in Table 1.

[0154]

[0155] Table 1 shows that, compared to the undamaged state of the reference section, the mode value of the strain ratio at some measuring points changes under Type I damage, indicating that the mode value can be used to diagnose crack damage. The stiffness ratio matrices of the mid-span U-rib section and the DP3 section of the diaphragm in the Span 3 segment under Type I damage are as follows:

[0156] (11)

[0157] (12)

[0158] From equations (12) and (13), it can be seen that, compared with the stiffness ratio distribution of other measuring points, β Ⅰ-U3 and β Ⅰ-D3 The changes were significant, with the largest absolute values ​​being 9% and 16%, respectively. This indicates that damage exists in the test section below the third U-rib and near the third strain measurement point on the diaphragm. Figure 8 The location of the crack damage on the steel box girder is consistent with the information provided.

[0159] Based on the strain ratio probability density distribution at each monitoring point, Figure 4 Diagnostic analysis was performed on Type II and Type III fatigue crack damage in the steel box girder. The location distribution of Type II and Type III fatigue crack damage is shown in the figure below. Figure 11As shown in the figure. Type II crack is located between the second and third U-ribs in section DP3, while Type III crack occurs at the mid-span of the second U-rib in section Span 3. The length, width, and depth of Type II and Type III cracks are 120 mm, 2 mm, and 10 mm, respectively. The strain ratio distribution at each measuring point of the reference section and the test section when Type II crack occurs is shown in the figure. Figure 12-13 As shown. By Figure 12-13 It can be seen that the strain ratio at the strain measurement point at the bottom of the U-rib did not change significantly, while the strain ratio distribution at the DP3-3 measurement point of the transverse diaphragm was significantly shifted compared to the reference section DP2-3. This indicates that the transverse diaphragm was damaged near the DP3-3 measurement point. Figure 11 The location of the Type II fatigue crack is consistent with that of the reference section. Compared with the reference section, the stiffness ratio matrices of the mid-span U-rib section and the diaphragm DP3 section of the Span 3 segment under Type II damage are shown in Equations (13) and (14), respectively. Compared with Type I damage, Type II damage only affects the stiffness ratio matrix of the diaphragm of the steel box girder.

[0160] (13);

[0161] (14).

[0162] The strain ratio distribution at each measuring point of the reference section and the test section when a type III crack occurs is as follows: Figure 14-15 As shown. By Figure 14-15 It can be seen that the strain ratios at each measuring point on the diaphragm did not change significantly, while the strain ratio distribution at the Span 3-U-2 measuring point at the bottom of the U-rib changed significantly compared to the Span 2-U-2 measuring point at the reference section. This indicates that the U-rib was damaged near the Span 3-U-2 measuring point. Figure 11 The location of the Type III fatigue crack is consistent with that of the reference section. Compared with the reference section, the stiffness ratio matrices of the U-rib section at the mid-span of the Span3 section and the DP3 section of the transverse diaphragm under Type III damage are shown in Equations (15) and (16), respectively. Compared with Type I damage, Type III damage only affects the stiffness ratio matrix of the U-rib of the steel box girder.

[0163] (15);

[0164] (16).

[0165] The above results indicate that Figure 4 The three different types of crack damage conditions have different effects on the stiffness ratio of steel box girders. Based on the different sensitivities of the strain ratio distribution of U-ribs and transverse diaphragms to different crack damage types, the locations of three typical local crack damages in steel box girders can be accurately diagnosed. Specific Implementation Example 2:

[0167] Combination Figure 1and Figures 16 to 19 Based on specific embodiment 1, this embodiment further provides a method for the coordinated assessment of local and overall fatigue damage of steel box girders for long-span bridges. The main contents are as follows:

[0168] by Figure 4 Taking Type I, II, and III crack damage as examples, the distribution law of stiffness ratio of each section of steel box girder under different crack lengths is solved by numerical simulation method. Specifically, the crack length gradually increases from 50 mm to 200 mm in increments of 0.5 mm, with crack width and depth of 2 mm and 10 mm, respectively. The location of Type I cracks is shown in the figure. Figure 8 As shown, the locations of type II and type III cracks are as follows: Figure 11 As shown.

[0169] Figure 16 The stiffness ratio distribution of the mid-span U-rib section in the Span 3 segment under different crack lengths for Type III crack damage is shown. Figure 16 It is evident that as the crack length increases, different wheel-load strain distributions appear at the strain monitoring points at the bottom of the U-ribs at mid-span of the Span 3 section, leading to varying local stiffness changes in the six U-ribs of the steel box girder's top plate. These changes are related to the characteristic parameters of bridge cracks along the U-ribs. Therefore, the stiffness ratio data from different monitoring points can be processed to effectively characterize crack type, length, and other characteristic parameters based on the different sensitivities of each measuring point to crack damage, thereby enabling performance evaluation and safety early warning for the steel box girder.

[0170] Fatigue crack damage diagnosis of steel box girders is a small-sample nonlinear multi-class classification problem. SVM (Simultaneous Dynamical Vector Machine) can transform the nonlinear curves of low-dimensional sample data to a high-dimensional hyperplane using a kernel function to solve multidimensional and nonlinear problems. The computational accuracy of the SVM model is affected by the penalty factor C and the kernel function parameter g. Traditional SVM models mainly rely on experience to obtain the optimal parameters, which lacks universality. In this embodiment, the GA (Generalized Algorithm) algorithm is used to find the optimal parameters C and g of the SVM. The technical route of the GA-SVM algorithm model is as follows: Figure 17 As shown.

[0171] Figure 16 The stiffness ratio distribution data in the model exhibits relatively small differences and is close to 1, with the difference between the maximum and minimum values ​​not exceeding 0.15. This high similarity among data features poses a challenge to SVM classification tasks. To improve the classification performance of SVM, the original stiffness ratio data needs to be processed to amplify the differences between data features. Nonlinear transformations can map the data to a new space by applying a nonlinear function to the original data, thereby amplifying the differences between data. This embodiment uses the Box-Cox transformation as a specific method for nonlinear transformation. The expression for the Box-Cox transformation is as follows:

[0172] (17)

[0173] In the formula, x is an observation in the original dataset; y(λ) is the new observation after the Box-Cox transformation; λ is a power parameter, and different values ​​of λ will produce different transformation effects.

[0174] After performing a nonlinear transformation on the original stiffness ratio data, the resulting new data may have different value ranges and dimensions. Without standardization, features with larger value ranges will dominate the SVM model, while features with smaller value ranges may be ignored. Standardization ensures all data features have the same scale, thus avoiding the impact of different data feature scales on the performance of the GA-SVM algorithm. This embodiment uses the z-score standardization method, and the data after the nonlinear transformation is... The standardized data is y, and its expression is:

[0175] (18)

[0176] In the formula, μ is the mean of the stiffness ratio data; σ is the standard deviation of the stiffness ratio data.

[0177] To further address the issue of small differences in overall performance parameters caused by local fatigue cracking damage in steel box girders, which leads to underfitting in the safety performance assessment model—meaning the model cannot capture complex patterns in the data and cannot effectively evaluate real-time data—this embodiment, based on the aforementioned Box-Cox transformation and standardization, further employs a binomial feature expansion method. By increasing the complexity of data features, this allows the SVM model to learn more complex nonlinear relationships, thereby improving the accuracy of the SVM model.

[0178] This embodiment uses Figure 4 Taking three types of crack damage as examples, Figure 18 A framework for fatigue damage assessment of steel box girders based on the GA-SVM algorithm model is presented. To effectively utilize the strain monitoring data from each measuring point, the established framework mainly includes five steps: First, numerical simulation is performed to obtain strain data from each measuring point. Then, the strain ratio and stiffness ratio distributions of the steel box girder under different damage conditions are obtained. Second, the obtained raw stiffness ratio feature data undergoes differential processing: first, a nonlinear transformation is performed on the raw data to amplify the data differences; then, standardization is performed to make the features have the same scale; finally, binomial feature expansion is performed to increase the complexity of the data features. The data processed in this way is more suitable for the classification task of the SVM model. Finally, the GA-SVM algorithm is used to obtain the relationship between the local cracking damage type of the steel box girder, its corresponding maintenance measures, and the strain ratio and stiffness ratio parameters of each measuring point, which is then used for real-time assessment of the safety status of the steel box girder.

[0179] It should be noted that those skilled in the art can construct a mapping model between stiffness ratio and crack length, crack propagation rate, and crack tip position based on indoor crack monitoring tests or field crack monitoring tests. This embodiment is only for verifying the effectiveness of the method of the present invention; therefore, a simulation method is used instead of indoor or field crack monitoring tests. That is, by artificially setting the crack length development, the strain response of each monitoring point on the test section under different crack lengths is simulated to obtain the corresponding stiffness ratio data. Based on this, a mapping relationship between stiffness ratio and crack length, crack propagation rate, and crack tip position is established.

[0180] In this embodiment, 903 sets of strain data were obtained through numerical simulation. Figure 4 The strain data for the three types of fatigue crack damage were collected in 301 sets. Based on the strain data of each monitoring point, the stiffness ratio data of the steel box girder was obtained, and the mapping relationship between the stiffness ratio distribution and the crack length was established. Thus, the value in equation (9) was determined. The value of is determined based on the following assumptions: when the crack length is less than 50 mm, the crack does not propagate; when the crack length is between 50 mm and 150 mm, the crack propagation rate is less than 5 mm / month; when the crack length is greater than 150 mm, the crack propagation rate is greater than 5 mm / month. The values ​​for the crack propagation rate are also 903 sets; assuming the crack tip position corresponds to... The value can be any of 903 random integers from 1 to 3. In this embodiment, 70% of the data samples are used as the training set and 30% as the test set. The input sample data and its quality will affect the accuracy of the GA-SVM model. In this embodiment, the Latin Hypercube (LHS) sampling method is used to construct the data sample set, sampling from two dimensions: strain measurement points on the diaphragm and strain measurement points at the bottom of the U-rib, to ensure uniform distribution and comprehensive coverage of the samples in the multidimensional parameter space.

[0181] After assigning numerical labels to fatigue crack types and fatigue crack safety levels, the output feature vector Q=[T, D] of the GA-SVM is constructed. The numerical labels for fatigue crack type T are set as follows: Type I crack 1, Type II crack 2, Type III crack 3; and the numerical labels for fatigue crack safety level D are set as follows: Level A 4, Level B 5, Level C 6. The GA algorithm is used to optimize the SVM model parameters C and g. Based on the actual debugging process, the final GA algorithm parameters are set as follows: population size 20, crossover probability 0.9, mutation probability 0.8, and termination generation 100. After 100 iterations, the optimal curve of the GA-SVM parameters gradually converges, and the optimal parameters of the SVM model can be obtained. Specifically, the penalty factor C=90.9531, and the kernel parameter g=0.0044.

[0182] In this embodiment, the accuracy of fatigue crack type diagnosis and fatigue crack safety level classification is used as the evaluation index of the proposed GA-SVM algorithm model. The calculation formula for the accuracy of the proposed GA-SVM algorithm model can be expressed as:

[0183] (19)

[0184] In the formula, R is the classification accuracy; P is the number of correctly classified samples; and S is the total number of samples.

[0185] Figure 19 This embodiment uses the GA-SVM algorithm model to diagnose local fatigue crack types and assess the overall structural safety level. To visually compare the impact of data differential expansion processing on the classification performance of the GA-SVM algorithm model, two scenarios are set up: Case 1 is the original stiffness ratio data without differential expansion processing, and Case 2 is the data after differential expansion processing. Regarding crack type diagnosis, since there are only three types of fatigue cracks in this embodiment, and different types of cracks have significant effects on the strain ratio distribution of the steel box girder structure—namely, Type I cracks affect the strain ratio distribution of both the diaphragms and U-ribs, Type II cracks primarily affect the strain ratio distribution of the diaphragms, and Type III cracks only affect the strain ratio distribution of the U-ribs—the GA-SVM algorithm model can accurately diagnose crack types with 100% accuracy even without data differential expansion processing. However, due to the small differences in the stiffness ratio distribution among the input monitoring points, the GA-SVM algorithm model only achieved a classification accuracy of 59.4% for the overall structural safety level under the training data distribution in Case 1, which is unfavorable for the safety assessment of steel box girders. In Case 2, after enhancing the data features through differential expansion, the GA-SVM algorithm model achieved an accuracy of 94.8% for the overall structural safety level. Differential expansion can significantly improve the accuracy of the overall structural safety level assessment for steel box girders.

[0186] Before using the aforementioned GA-SVM algorithm model to conduct a fatigue damage safety assessment of a real bridge, the GA-SVM algorithm model needs to be trained based on stiffness ratio data of fatigue crack types observed in other existing steel box girder bridges at different service times. The model should encompass as many possible damage conditions as possible to achieve higher assessment accuracy (the error should generally be controlled within 5%). Once the training results meet expectations, the obtained GA-SVM algorithm model is used to assess the real bridge, enabling one-time measurement and assessment. Only the strain response at each monitoring point under any service time needs to be obtained through sensors, eliminating the need for long-term direct monitoring of cracks on the real bridge, thus saving a large amount of repetitive monitoring work that is susceptible to environmental interference. When fatigue damage is assessed, direct monitoring methods (such as manual inspection) or even maintenance measures are then used to address the cracks.

[0187] The method of this invention uses strain ratio and stiffness ratio as the criteria for judging fatigue damage, and coordinates the assessment of local fatigue damage with the overall structural safety of the steel box girder. This allows for a more accurate diagnosis and assessment of crack development, effectively avoiding the blindness of on-site testing, saving time and economic costs, and ensuring the overall safety of the steel box girder. For other types of local damage conditions, the technical concept of this invention can also be applied to propose corresponding assessment methods that coordinate local fatigue damage with the overall safety of the steel box girder.

[0188] Implementation method two.

[0189] This embodiment provides a system for the coordinated assessment of local and overall fatigue damage in steel box girders of long-span bridges, used to perform a method for the coordinated assessment of local and overall fatigue damage in steel box girders of long-span bridges, including:

[0190] The data acquisition module is used to: collect the strain response of each monitoring point on the reference section and test section of the steel box girder before and during the current service time in real time;

[0191] The data processing module is used to: calculate the stiffness ratio of each monitoring point on the test section at the current service time based on the strain response; and use the GA-SVM algorithm model to obtain the fatigue crack type and fatigue crack safety level of the steel box girder based on the differentially processed stiffness ratio data.

[0192] The data storage module is used to store monitoring data, data during data processing, and data on fatigue crack types and fatigue crack safety levels of steel box girders.

[0193] The data output module is used to output the fatigue crack type based on the stiffness ratio of the current service time of each monitoring point on the test section, as well as the fatigue crack safety level and countermeasures for the steel box girder.

[0194] Furthermore, based on the fatigue crack safety level, the following countermeasures should be taken:

[0195] When the safety level is A, the response measure is to conduct regular observation of the cracks at least once every six months;

[0196] When the safety level is B, the response is to closely observe the cracks at least once a month.

[0197] When the safety level is C, the response is to promptly repair and reinforce the cracks as necessary.

[0198] When the safety level of a steel box girder is assessed as Grade A or B during its current service life, regular or close observation refers to the process where the method of this invention has diagnosed localized fatigue damage in the steel box girder. In this case, manual or machine inspections can be initiated to promptly observe the crack development under normal vehicle loads and intervene in the crack development in a timely manner. If the safety level of a steel box girder is assessed as Grade C during its current service life, and it is determined that the steel box girder requires repair and reinforcement, the repair and reinforcement methods are generally determined based on factors such as the crack's location, length, tip position, and propagation path. For example, for cracks with a length not exceeding 150mm, the pneumatic impact method is often used (the pneumatic impact method uses high-speed impact from pneumatic tools to cause plastic deformation of the surface metal near the crack opening, forming crack closure, and simultaneously introducing residual compressive stress to ensure tight crack closure). When the crack tip is on the base material of the component and the length does not exceed 300mm, drilling can be used for repair along the fatigue crack propagation path. After repairing and reinforcing the crack damage, it is necessary to conduct a dynamic load test on the bridge as needed. If a dynamic load test is carried out, the data obtained from this load test will be the starting point for a new round of strain ratio calculation.

[0199] It should be noted that the raw monitoring data from the sensor are typically discrete points that change over time, while the strain response mentioned above refers to a continuously changing curve plotted from the discrete points monitored by the sensor. In this invention, the strain response can refer to either continuous discrete points or the corresponding continuous curve.

[0200] It should be noted that the data output module can output the stiffness ratio of monitoring points on the test section at the current service time, as well as the time history curve of the stiffness ratio of monitoring points near fatigue damage. These stiffness ratio parameters can be used as auxiliary information for the diagnosis of different fatigue crack damage locations. The data output module includes a display, which in some embodiments can be an LED display, an LCD display, a touch LCD display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display is used to display data information and a user interface for visualization. This user interface has display windows for the arrangement of monitoring points on the reference section and the test section, display windows for the strain response of different monitoring points, display windows for the strain ratio histogram of different monitoring points, and display windows for crack level and safety level. It can provide early warning prompts based on the output safety level, such as flashing display windows or text prompts, and can even be connected to external buzzers or flashing lights to remind personnel to take relevant countermeasures.

[0201] It should be noted that the data storage module uses a memory 20. In some embodiments, the memory 20 can be an internal storage unit of the terminal, such as the terminal's hard drive or memory, or an external storage device of the terminal, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. It can also include both internal and external storage units, or it can be cloud storage. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory can also be used to temporarily store data that has been output or will be output. In this embodiment, the memory stores a calculation program for a method for the coordinated evaluation of local and overall fatigue damage of steel box girders in long-span bridges, which can be executed by the data processing module.

[0202] It should be noted that the data processing is implemented on a computing platform, which can be a microcomputer, a laptop, or other computing platform with computing capabilities. Processing software, such as MATLAB, is installed on the computing platform. The data processing program for the local and overall fatigue damage co-assessment method for steel box girders proposed in this invention is incorporated into the MATLAB or other software on the computing platform. According to the method in Embodiment 1, those skilled in the art can write the corresponding calculation program in the software, which will not be elaborated further here.

[0203] The above embodiments are only for illustrating the technical concept and features of the present invention. Their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be used to limit the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for local and global collaborative evaluation of fatigue damage of a long-span bridge steel box girder, characterized in that: The method comprises the following steps: Step 1: selecting a reference section and a test section on the steel box girder, and arranging a plurality of strain sensors as monitoring points at corresponding positions on the two selected sections; Step 2: before the steel box girder is in service, carrying out a bridge dynamic load test on the steel box girder, collecting strain responses of the monitoring points on the two selected sections in real time through the sensors; and aligning the strains on the two selected sections according to response times by using a cross-correlation method; Step 3: defining a strain ratio of any monitoring point on the test section relative to a corresponding monitoring point on the reference section at a certain time as a strain ratio; calculating the strain ratios of any monitoring point on the test section relative to the corresponding monitoring point on the reference section at different times according to the aligned strain responses in Step 2; and obtaining statistical characteristic values of the strain ratios of the any monitoring point before the service by using a data statistical method; Step 4: after the steel box girder is in service, carrying out a bridge dynamic load test on the steel box girder at a current service time, and obtaining statistical characteristic values of the strain ratios of the any monitoring point at the current service time by using the method in Steps 2 and 3; Step 5: defining a stiffness ratio of the any monitoring point on the test section of the steel box girder at the current service time relative to the strain ratio before the service as a stiffness ratio; calculating the stiffness ratio of the any monitoring point on the test section at the current service time according to the statistical characteristic values of the two types of strain ratios obtained in Steps 3-4; taking a position with the largest absolute value change of the stiffness ratio as a crack damage position according to a distribution of the stiffness ratios on the same section; and determining a crack damage type according to the damage position; Step 6: differentiating the stiffness ratio obtained in Step 5, and obtaining different fatigue crack damage types of the steel box girder and corresponding fatigue crack safety levels by using a genetic algorithm to optimize a support vector machine model according to the differentiated stiffness ratio data.

2. The method for local and global collaborative evaluation of fatigue damage of long-span bridge steel box girder according to claim 1, characterized in that: In Step 1, the test section is a section within 1 / 3-2 / 3 of the span direction of the steel box girder, and the reference section is any section within 1 / 5-1 / 3 or 2 / 3-4 / 5 of the span direction of the steel box girder.

3. The method for local and global collaborative evaluation of fatigue damage of long-span bridge steel box girder according to claim 1, characterized in that: In Step 1, the fatigue crack damage of the steel box girder includes three types: type I damage is transverse cracking at a welding position between a transverse diaphragm and a U rib and at an opening position of the transverse diaphragm, type II damage is transverse cracking at a welding position between the transverse diaphragm and a top plate, and type III damage is longitudinal cracking at a welding position between the U rib and the top plate; and strain monitoring points are arranged at the U rib bottom in the span direction and at the opening position of the transverse diaphragm in the vertical bridge direction.

4. The method for local and global collaborative evaluation of fatigue damage of long-span bridge steel box girder according to claim 1, characterized in that: In Steps 3-4, the strain ratio is obtained according to formulas (3.1) and (4.1): (3.1) (4.1) In the formula, the reference section the measured strain value of the upper monitoring point i at a certain time; the test section before the steel box girder is in service the measured strain value of the upper monitoring point i at a certain time; the test section of the steel box girder at the current service time T the measured strain value of the upper monitoring point i at a certain time; the strain ratio of the upper monitoring point i on the test section of the steel box girder before service to the upper monitoring point i on the reference section; the strain ratio of the upper monitoring point i on the test section of the steel box girder at the current service time T to the upper monitoring point i on the reference section; the distance of a certain section of the steel box girder from the end section.

5. The method for local and global collaborative assessment of fatigue damage of long-span bridge steel box girder according to claim 1, characterized in that: In Step 3, the data statistical method uses a probability density function; and the statistical characteristic value uses one of a mode value, a maximum value and a mean value.

6. The method for local and global collaborative assessment of fatigue damage of long-span bridge steel box girder according to claim 1, characterized in that: In Step 5, the stiffness ratio is obtained according to formula (5.1): (5.1) In the formula, Test section for current service time T of steel box girder Stiffness ratio of upper monitoring point i.

7. The method for local and global collaborative assessment of fatigue damage of long-span bridge steel box girder according to claim 1, characterized in that: In Step 6, the differentiation processing is: firstly, performing nonlinear transformation on the original stiffness ratio data; secondly, performing standardization processing on the nonlinearly transformed stiffness ratio data; and finally, performing feature expansion on the standardized stiffness ratio data through a polynomial to obtain differentiated stiffness ratio data.

8. The method for local and global collaborative assessment of fatigue damage of long-span bridge steel box girder according to claim 1, characterized in that: In Step 6, the process of solving the fatigue crack safety level of the steel box girder by using the genetic algorithm to optimize the support vector machine model comprises the following steps: 1) According to the stiffness ratio data under the current service time, it is substituted into the mapping model of stiffness ratio and crack length, crack propagation rate, crack tip position to obtain the crack length, crack propagation rate and crack tip position under the current service time; The mapping model is fitted according to the indoor crack monitoring test or field crack monitoring test, and is established according to formula (6.1): (6.1) wherein, , , Lj, Vj, Xjare the length, the propagation rate, the tip position of the crack j at the current service time T, respectively; , , are the mapping relationships of the stiffness ratio and the crack length, the crack propagation rate, the crack tip position, respectively; and Φ is a comprehensive function. 2) According to the crack length, crack propagation rate and crack tip position under the current service time, the fatigue crack safety factor under the current service time is obtained according to formula (6.2): (6.2) wherein is the fatigue crack safety factor; , , is the current service time the length, the propagation rate, and the tip position of the crack j. wherein, The score of the item is determined by the following method: when the crack length is greater than 0 and less than 100 mm, = 1; when the crack length is greater than or equal to 100 mm and less than 150 mm, = 2; when the crack length is greater than or equal to 150 mm, = 3; wherein, The score of the item is determined as follows: when the crack does not propagate, = 1 ; when the crack propagation rate is greater than 0 and less than 5 mm / month, = 2; when the crack propagation rate is greater than 5 mm / month, = 3; in, The score is determined as follows: when the crack tip is located at the weld toe or weld root, =1; when the crack tip is within the weld height. =2; when the crack tip is located in the base material. =3; 3) According to the fatigue crack safety factor under the current service time, the fatigue crack safety grade is output; The fatigue crack safety grade of the steel box girder is evaluated according to the following method: When 0.3≤λ<0.5, the safety grade is A level; When 0.5≤λ<0.7, the safety grade is B level; When 0.7≤λ≤1.0, the safety grade is C level. 9.A system for local and global collaborative evaluation of fatigue damage of a long-span bridge steel box girder, configured to perform the method for local and global collaborative evaluation of fatigue damage of a long-span bridge steel box girder according to any one of claims 1-8, characterized in that: It comprises: A data acquisition module for acquiring the strain response of each monitoring point on the reference section and test section of the steel box girder before and at the current service time in real time; A data processing module for calculating the stiffness ratio of each monitoring point on the test section at the current service time according to the strain response, and obtaining the fatigue crack type and fatigue crack safety grade of the steel box girder by using the GA-SVM algorithm model according to the differentiated stiffness ratio data; A data storage module for storing monitoring data, data during data processing and data of the fatigue crack type and fatigue crack safety grade of the steel box girder; A data output module for outputting the fatigue crack type of the steel box girder and the corresponding fatigue crack safety grade and countermeasures.

10. The long-span bridge steel box girder fatigue damage local and global collaborative evaluation system according to claim 9, characterized in that: According to the fatigue crack safety grade, the countermeasures are taken according to the following method: When the safety grade is A level, the countermeasures are to observe the crack at least once every half year; When the safety grade is B level, the countermeasures are to closely observe the crack at least once every month; When the safety grade is C level, the countermeasures are to timely repair and reinforce the damage of the crack.

Images