Welding node residual fatigue life evaluation method based on multi-source heterogeneous information fusion

Abstract

The invention discloses a welding node residual fatigue life evaluation method based on multi-source heterogeneous information fusion. The method comprises the steps that sensitivity analysis is conducted on uncertainty factors in the fatigue crack initiation and expansion process with the crack length as the judgment quantity; according to the analysis result, based on the Bayesian theory, a railway steel bridge welding joint degradation model is established; then acquiring a crack length image, an acoustic emission signal and a strain signal; inputting into a deep learning neural network to obtain a crack length; and inputting the crack length into a railway steel bridge welding joint degradation model to obtain the residual fatigue life of the welding spot. According to the method, the uncertainty of monitoring data, material mechanical properties and a theoretical model can be comprehensively considered, so that the prediction precision of the residual life of the steel bridge welding spot is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fatigue life prediction and evaluation of railway steel bridge welded nodes, and more particularly to a method for evaluating the remaining fatigue life of welded nodes based on multi-source heterogeneous information fusion. Background Art

[0002] Fatigue cracking has long been a prominent issue in the design and maintenance of railway steel bridges. Short fatigue cracks or long fatigue cracks extending through the thickness of welded joints are frequently detected, posing a serious safety hazard to railway bridge operations. Therefore, accurately determining the current crack status at critical welded joints and assessing their growth trends and remaining lifespan are of great research significance and engineering value for ensuring the safe operation of railway steel bridges.

[0003] Current remaining life assessment methods are mainly divided into fatigue test-based methods and fracture mechanics theory-based methods;

[0004] The former primarily relies on fatigue testing, which estimates the remaining life of welded joints based on experimentally measured SN curves. Fatigue test data exhibit significant random discreteness, requiring probabilistic analysis based on a sufficient amount of test data to identify statistical patterns. Due to limitations in R&D cycles and funding, the number of fatigue test specimens used in various studies is very limited. The SN curves obtained must be verified to determine whether they meet specific requirements, and the research results lack a clear understanding of causal relationships. Therefore, fatigue testing-based methods struggle to accurately determine the remaining fatigue life of railway steel bridges.

[0005] The latter method uses fracture mechanics to analyze and predict the growth of long fatigue cracks in critical weld joints. While this method avoids the time and cost associated with fatigue testing, it cannot accurately predict the low-stress-amplitude fatigue life, which accounts for a significant portion of railway steel bridge service life, because fracture mechanics cannot assess the life span of fatigue crack initiation and short crack initiation stages. Furthermore, the complex service environment of railway steel bridges often affects the growth of long fatigue cracks due to a variety of uncertainties, resulting in low accuracy in long fatigue crack growth predictions.

[0006] Therefore, fracture mechanics theory cannot be directly applied to the accurate value of the remaining fatigue life of railway steel bridges. Summary of the Invention

[0007] In view of this, the present invention integrates information from multiple information sources of the service health monitoring system of serving steel bridges, and based on the fatigue full life theory, establishes a real-time accurate value evaluation method for the remaining life of welding nodes to comprehensively consider the uncertainty of monitoring data, material mechanical properties and theoretical models.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] A method for residual fatigue life assessment of welded joints based on multi-source heterogeneous information fusion, including:

[0010] A sensitivity analysis of the uncertainty factors in the fatigue crack initiation and propagation process was conducted using crack length as a criterion. Based on the analysis results and the Bayesian theory, a railway steel bridge weld node degradation model was established.

[0011] Crack length images, acoustic emission signals, and strain signals are collected and input into a deep learning neural network to obtain the crack length. The crack length is then input into a railway steel bridge weld node degradation model to obtain the residual fatigue life of the weld.

[0012] Specifically, the method for evaluating the remaining fatigue life of welded joints based on multi-source heterogeneous information fusion disclosed in the present invention has the following beneficial effects compared with the prior art:

[0013] 1) A high-precision fatigue crack identification method based on deep learning neural networks was established, significantly improving the efficiency and accuracy of fatigue crack identification in railway steel bridges;

[0014] 2) To simulate the fatigue damage evolution of welded joints in railway steel bridges subjected to random loads under time series, a dynamic Bayesian network model for fatigue crack initiation and propagation was established. The uncertainties in the fatigue crack initiation and propagation processes were used as state nodes in the model to accurately estimate the fatigue crack size and remaining fatigue life under random loads.

[0015] 3) To reduce the impact of uncertainty factors on the assessment accuracy during fatigue crack initiation and growth, an uncertainty parameter inference algorithm based on instrumented indentation and particle filter inference was established. Using the key material data of fatigue-prone areas of welded joints collected by instrumented indentation, the impact of multiple uncertainty factors on the fatigue crack initiation and growth process was gradually reduced.

[0016] 4) Based on the real fatigue crack data collected by multiple sensors, the uncertainty parameters in the fatigue crack initiation and propagation process are gradually updated to achieve dynamic tracking of fatigue crack propagation and real-time and accurate evaluation of the remaining fatigue life. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0018] Figure 1 This is a flow chart of the residual fatigue life assessment method for welded joints based on multi-source heterogeneous information fusion;

[0019] Figure 2 Flowchart for solving cyclic constitutive curves;

[0020] Figure 3 Solve the flow chart for the circulation resistance curve;

[0021] Figure 4 Flowchart for solving crack growth rate curve;

[0022] Figure 5 Flowchart of the multi-source information crack size recognition algorithm based on deep learning. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0025] To address the shortcomings of existing crack life assessment methods, this application adopts deep learning, dynamic Bayesian network and nonlinear prediction theory of fatigue short cracks.

[0026] Specifically, in order to reduce the impact of uncertainty factors on the residual fatigue life of railway steel bridge welded joints during the evaluation process, this embodiment provides a method for evaluating the residual fatigue life of welded joints based on multi-source information fusion, such as Figure 1 ,include:

[0027] A sensitivity analysis of the uncertainty factors in the fatigue crack initiation and propagation process was conducted using crack length as a criterion. Based on the analysis results and the Bayesian theory, a railway steel bridge weld node degradation model was established.

[0028] Crack length images, acoustic emission signals, and strain signals are collected and input into a deep learning neural network to obtain the crack length. The crack length is then input into a railway steel bridge weld node degradation model to obtain the residual fatigue life of the weld.

[0029] Example 1: Establishing a degradation model for railway steel bridge welded joints

[0030] This example uses fatigue cracks under variable amplitude loads as a specific object and uses the local stress-strain method, the IBESS method, and a high-precision fatigue long crack propagation prediction method to construct a physical state model of fatigue crack initiation, short crack propagation, and long crack propagation to characterize the multi-stage degradation process of fatigue crack initiation, short crack propagation, and long crack propagation in welded joints under variable amplitude loads. The relationship between typical uncertainties in the fatigue crack initiation and propagation process is then analyzed, and a dynamic performance degradation model for fatigue crack initiation and propagation is established based on a dynamic Bayesian network. The steps include:

[0031] Step 1: Use crack length as a criterion to conduct sensitivity analysis on the uncertainty factors in the fatigue crack initiation and propagation process; including:

[0032] When the crack length is less than 0.1 mm, the crack is in the crack initiation stage, and the fatigue life is solved using the local stress-strain method;

[0033] When the crack length is between 0.1 mm and 0.5 mm, the IBESS method is used to solve the fatigue life;

[0034] When the crack length is greater than 0.5 mm, the long crack growth prediction algorithm is used to solve the fatigue life.

[0035] 1) When solving fatigue life using the local stress-strain method, the measured load spectrum and weld size are required. Based on structural mechanics knowledge, the nominal stress spectrum at the fatigue crack of the railway steel bridge weld node is obtained. The local elastic-plastic stress spectrum data at the crack is obtained by combining the cyclic constitutive curve of the steel and the Neuber criterion. The residual fatigue life is calculated by combining the cumulative damage criterion. The relevant formula is as follows:

[0036] The fatigue crack initiation life of the welded joint in the two-stage model is calculated using the strain-fatigue life formula for smooth specimens. The Basquin-Coffin-Manson formula describes the strain-fatigue life of smooth specimens and can be expressed as:

[0037]

[0038] Where: Δε is the plastic strain amplitude, E is the elastic modulus, N i is the fatigue crack initiation life, ε' f is the fatigue ductility coefficient, σ' f is the fatigue strength coefficient, c is the fatigue ductility index, and b is the fatigue strength index. When considering the influence of the average stress on the strain-fatigue life of the smooth specimen, it is necessary to correct the formula (1), and the corrected formula is:

[0039]

[0040] Where: σ m is the mean stress, and the other symbols are the same as those in formula (1).

[0041] When steel is subjected to cyclic loading, its stress-strain relationship can be given by the Ramberg-Osgood formula, which is:

[0042]

[0043] Where Δε is the strain amplitude, Δσ is the stress amplitude, E is the elastic modulus, K′ is the cyclic strength coefficient, and n′ is the cyclic strain hardening coefficient.

[0044] There is stress concentration at the notch root, and its stress is not directly equal to the far-field stress. Topper et al. [modified the Neuber criterion and found that the stress amplitude and strain amplitude of the notch root cycle and the far-field stress amplitude satisfy the following relationship:

[0045]

[0046] Where: Δε is the notch strain amplitude, Δσ is the notch stress amplitude, E is the elastic modulus, K f is the fatigue notch factor, S is the far-field stress amplitude. Fatigue notch factor K f It refers to the ratio of the fatigue strength of the smooth specimen to the fatigue strength of the notched specimen.

[0047] 2) When solving fatigue life, the IBESS method considers the expansion process of short and long cracks at the same time. Similarly, the internal force spectrum is obtained based on the measured load spectrum and weld size, according to the knowledge of structural mechanics, and then the stress intensity factor spectrum is obtained based on the knowledge of fracture mechanics. The plastic correction stress intensity factor amplitude at the fatigue crack is obtained by combining the crack closure parameter considering the crack closure effect and the cyclic resistance curve of the short crack threshold. Combined with the crack growth curve, the remaining life is obtained according to the "Cycle-by-Cycle" algorithm. The relevant formulas include:

[0048] The expression of crack growth rate of short crack is:

[0049]

[0050] Where: da / dN is the crack growth rate; ΔK is the total driving force applied to the crack; ΔK th is the crack extension threshold; C and m are constants determined by the material.

[0051] The effective critical value of long crack ΔK th,eff Instead of the microstructure critical value ΔK dR The Chapetti model expression is used to estimate the threshold ΔK for short crack growth th :

[0052] ΔK th =ΔK dR +(ΔK thR -ΔK dR )[1-e -k(a-d) ] (6)

[0053] Where: ΔK thR is the threshold value of the stress intensity factor amplitude of a long crack; a is the crack length, and k is the material constant given by the following expression:

[0054]

[0055] Where: d is the distance from the surface of the strongest microstructural barrier; ΔK thR is the threshold value of the stress intensity factor amplitude of the long crack; ΔK dR is the microstructural threshold for crack growth.

[0056] 3) When calculating fatigue life using the long crack growth prediction algorithm, the internal force spectrum is obtained based on the measured load spectrum and weld size, according to structural mechanics knowledge. The stress intensity factor spectrum is then obtained based on fracture mechanics knowledge, and the effective stress intensity factor spectrum is obtained based on the crack closure parameters. The remaining life is then calculated using the cycle-by-cycle algorithm based on the effective stress intensity factor spectrum combined with the crack growth curve. The relevant formula is as follows:

[0057] The SIF formula for a plate with a semi-elliptical surface crack can be expressed using the following hybrid formula:

[0058]

[0059] Where: K is stress intensity factor (SIF); F E is the crack shape correction factor; F S is the crack front correction factor; F T is the finite thickness correction factor; F W is the finite width correction factor; F D is the stress gradient correction factor; σ is the stress at the crack tip; a is the crack depth.

[0060] The expression of crack growth rate of long crack is:

[0061]

[0062] Where: da / dN is the crack growth rate; ΔK is the total driving force applied to the crack; ΔK th is the crack extension threshold; C and m are constants determined by the material.

[0063] Step 2: Based on the analysis results and the Bayesian theory, a railway steel bridge weld node degradation model is established. Specifically, the model includes setting high-sensitivity parameters as state nodes in the dynamic Bayesian network and low-sensitivity parameters as fixed nodes, and establishing a railway steel bridge weld node degradation model for crack initiation and propagation.

[0064] To further optimize the above technical solution, a particle filter inference algorithm is used to input the measured fatigue crack observation data into the dynamic performance degradation model, correct the prediction results, reduce the influence of uncertain factors, and obtain the real-time accurate value of the remaining fatigue life.

[0065] Specifically, the monotonic constitutive curve of the fatigue crack risk area of ​​the weld node is obtained based on the hardness method. Based on the monotonic constitutive curve, the initial cyclic constitutive curve is obtained by the Uniform Materials method, the initial cyclic resistance curve is obtained by the Leonetti method, and the initial crack growth curve is obtained by the iLAPS method.

[0066] The obtained initial curve is used as the initial parameter value of the particle filter algorithm. According to the principle of the particle filter algorithm, the posterior probability distribution of the uncertain node is represented by a random particle set with weights. By inputting the observed value into the model and comparing it with the estimated value, the particle set weight is iteratively updated to gradually approach the true posterior distribution, thereby reducing the impact of parameter uncertainty.

[0067] In an exemplary embodiment, the detailed steps are as follows:

[0068] (1) The instrumented indentation method is used to measure the hardness value HV of the fatigue cracking point of the railway steel bridge welding node. According to the hardness HV and the cyclic strength coefficient K' and the cyclic yield strength σ yc , cyclic strain hardening index ', to obtain these data, according to the steel cyclic constitutive curve formula, use the above data to obtain the initial cyclic constitutive curve, the steps refer to Figure 2 , the initial cyclic constitutive curve will be used for the local stress-strain method, and the above parameters can be used as the initial values ​​for the analysis.

[0069] For the initial cycle constitutive curve, this embodiment introduces the crack closure parameter U and proposes a new high cycle fatigue crack growth model in order to accurately and reliably give the median curve of high cycle fatigue crack growth rate for most metals. That is:

[0070]

[0071] Where: da / dN is the crack growth rate, U is the long crack closure parameter, c is the fatigue ductility index, E is the material elastic modulus, ε′ f Fatigue ductility coefficient, ΔK is the stress intensity factor amplitude, σ yc is the cyclic yield strength, n' is the cyclic strain hardening exponent, ΔK th is the stress intensity factor amplitude threshold.

[0072] The monotonic tensile curves of metal materials are different, so the corresponding cyclic stress-strain relationship should be determined by the following relationship, namely the Ramberg-Osgood equation:

[0073]

[0074] Where: Δε is the strain amplitude, Δσ is the stress amplitude, E is the elastic modulus, K′ is the cyclic strength coefficient, and n′ is the cyclic strain hardening exponent. Here, the definition method of monotonic yield strength can be used to calibrate the cyclic yield strength σ yc , the expression is as follows:

[0075] σ yc =K′(0.002) n′ (13)

[0076] The cyclic strain-life relationship can be described by the Coffin-Manson model, namely:

[0077]

[0078] Where: Δε is the plastic strain amplitude, Δσ is the stress amplitude, Ni is the fatigue crack initiation life, ε' f is the fatigue ductility coefficient, σ' f is the fatigue strength coefficient, c is the fatigue ductility index, and b is the fatigue strength index.

[0079] Muralidharan et al. proposed a correlation model that is comparable in accuracy to the general slope method, namely:

[0080]

[0081] Where: ε f is the fracture strain, E is the elastic modulus of the material, ε' f is the fatigue ductility coefficient, σ' fis the fatigue strength coefficient, c is the fatigue ductility index, and b is the fatigue strength index. %RA is defined as the section reduction rate, which is related to ε f The relationship is:

[0082]

[0083] The homogeneous material method can be called a slope method. The difference is that the estimated values ​​of the parameters used are significantly different in the two types of materials. The parameter ε' for titanium alloys and aluminum alloys is f It is considered to be a fixed value of 0.35; the parameter ε' of steel material f Then Related. That is:

[0084]

[0085] Where: ε f is the fracture strain, E is the elastic modulus of the material, ε' f is the fatigue ductility coefficient, σ' f is the fatigue strength coefficient, c is the fatigue ductility index, and b is the fatigue strength index.

[0086] Furthermore, in order to better use the four-point method model, the parameters are modified, namely:

[0087] σ' f =σ b (1+ε f ),ε' f =ε f

[0088]

[0089]

[0090] Where: ε f is the fracture strain, the monotonic tensile strength σ b , E is the elastic modulus of the material, ε' f is the fatigue ductility coefficient, σ' f is the fatigue strength coefficient, c is the fatigue ductility index, and b is the fatigue strength index.

[0091] At the same time, HB and tensile strength σ b The functional relationship between them is:

[0092]

[0093] Where: ε' f is the fatigue ductility coefficient, σ' f is the fatigue strength coefficient and HB is the hardness.

[0094] In the median method, Meggiolaro et al. proposed that LCF parameters can be estimated for aluminum alloys and steels, namely:

[0095] σ' f =1.5σ b , ε' f =0.45

[0096] b=-0.09,c=-0.59 (20)

[0097] Park et al. proposed that only the monotonic tensile strength σ b To estimate the relevant LCF model parameters, namely:

[0098]

[0099] Cyclic strain hardening exponent n′, Hu et al. found that it is related to 1-σ by mathematical transformation. yc / K′, that is:

[0100] n′=0.34(1-σ yc / K′)-0.05 (22)

[0101] Regarding the cyclic strength coefficient K′, some scholars proposed the following formula:

[0102] K′=57K 0.545 -1220 (23)

[0103] Where: K is the monotonic strength coefficient, K′ is the cyclic strength coefficient.

[0104] (2) The instrumented indentation method was used to measure the hardness value HV of the fatigue cracking site of the railway steel bridge welding node. According to the hardness HV and the defect size Intrinsic stress intensity factor amplitude threshold ΔK th，eff and maximum defect size The relationship between the parameters is used to obtain the data of these parameters. According to the empirical relationship between the cyclic resistance curve calibration parameters and the above parameters, the cyclic resistance curve calibration parameters are obtained. Combining the stress ratio at the weld node obtained by the load spectrum and structural mechanics knowledge, the long crack stress intensity factor amplitude threshold value in the existing technical specifications, and the cyclic resistance curve calibration parameter k, the stress intensity factor amplitude threshold value ΔK is obtained. th，eff With crack size The changing law of the initial circulation resistance curve is shown in the following steps. Figure 3 , the initial cycle constitutive curve will be used in the IBESS method, and the above parameters can be used as the initial values ​​for the analysis.

[0105] For the initial cycle resistance curve, in order to simulate the lower stress intensity factor amplitude threshold and higher crack growth rate of short cracks compared to long cracks, the stress intensity factor amplitude threshold and crack velocity function were modified to clarify the dependence on crack size. The accumulation model of external effects related to long cracks is as follows:

[0106] K th =ΔK th,eff +(ΔK th,lc -ΔK th,eff )[1-exp(-k(ad))] (24)

[0107] Where: ΔK th is the stress intensity factor amplitude threshold, ΔK th，eff is the intrinsic stress intensity factor amplitude threshold, ΔK th，lc is the threshold value of the stress intensity factor amplitude of the long crack, k is the calibration parameter of the cyclic resistance curve, a is the crack depth, and d is the distance from the free surface of the defect.

[0108] The parameter k controls the formation of the external effect and the length scale over which this occurs, and 1 / k represents the crack length at which the external effect, at some initial value, decreases to 1 / e of its initial value, where e is the Euler number. Chapetti uses k to estimate the distance d from the free surface of the defect, using the following formula:

[0109]

[0110] This relationship was derived by introducing micro-defects of different sizes into the polished specimens. For smaller defect sizes, the threshold stress range was found to be equal to the fatigue limit. This leads to The definition of the critical value of the parameter, that is, It indicates the maximum defect size that does not extend within the stress range below the fatigue limit. When the threshold stress coincides with the fatigue limit, the existing formula cannot be used. The hardness is used to estimate it. The formula is as follows:

[0111]

[0112] Where: Hv is the material hardness, the maximum defect size

[0113] Estimate the size of the non-extending crack, considering the defect to be semicircular, and Converted to crack depth, assuming a crit is a reasonable estimate of d, and a crit The calculation formula for d is as follows:

[0114]

[0115] (3) The instrumented indentation method is used to measure the hardness value HV of the fatigue cracking point of the railway steel bridge welding node. According to the hardness HV and the cyclic strength coefficient K′ and cyclic yield strength σ yc , fatigue ductility index n′, fatigue ductility coefficient c and cyclic strain hardening exponent n′, and obtain the above parameter values. The iLAPS model is used to obtain the initial crack growth rate curve. The steps refer to Figure 4 The initial cycle constitutive curve will be used in the IBESS method and long crack growth prediction algorithm, and the above parameters can be used as the initial values ​​in the analysis.

[0116] For the initial crack growth curve, according to the parameter σ yc , monotonic yield strength σ y and monotonic tensile strength σ b Scatter plot of the ratio, Fatemi et al. found that the parameter σ yc Existence The obvious segmented distribution of the boundary points. Based on the fitting of the data points, the segmented estimation formula of the cyclic plasticity parameters for the new model iLAPS is obtained:

[0117]

[0118] The cyclic strain hardening exponent n′ and the segmented cyclic yield strength σ yc There is a correlation with the parameter K′. Through some data analysis, the following expression is obtained:

[0119]

[0120] According to the experimental data processing results, the estimation error for some metal materials is too large. After refitting, the following results are obtained:

[0121] K′=8K 0.719 (30)

[0122] Referring to the calculation formula of Hu et al., the strain hardening coefficient n is calculated by the monotonic yield strength σ y and monotonic tensile strength σ b The estimate is:

[0123]

[0124] Cyclic yield strength σ yc It can be obtained from formula (28). Referring to formula (31), it can be known that n′ can be obtained by calculating the cyclic yield strength σ yc And the cyclic strength coefficient K′ is estimated, and the following formula is obtained through numerical analysis:

[0125]

[0126] Fatigue ductility coefficient ε' f It can be generally expressed by the nominal fracture strain and ε f To characterize, Genel believes that ε f and monotonic tensile strength σ b and the monotonic yield strength σ y There is a certain correlation between the ratio of ε' f It should also be considered The influence of the value of on itself, the following conclusions are obtained through fitting the experimental data:

[0127]

[0128] Some scholars have shown that the parameter c and ε f ,σ b It is related to the elastic modulus E. According to formula (18), the fatigue ductility index c can be obtained from the fracture strain and ε f , monotonic tensile strength σ b The ratio of the elastic modulus E is expressed as two variables, and it is also necessary to consider Correction of the final result and fitting parameters yielded the following results:

[0129]

[0130] The long crack closure parameter U can be defined as a function of the stress ratio R. This is because in the long crack stage, fatigue crack propagation is free from the microstructural barriers and is mainly related to the applied load. The crack closure function (UNC) is obtained by converting the Codrington closure function, as shown in the following formula:

[0131] U=0.52+0.35R+0.14R 2 (R≥-2) (35)

[0132] U=0.45+0.37R+0.20R 2 (R≥0) (36)

[0133] From the above two formulas, we can see that different calculation formulas are used under different stress ratios R. Therefore, the two functions are numerically fitted to obtain a new function (New-God) to obtain the following results:

[0134] U=0.49+0.35R+0.17R 2 (37)

[0135] Substitute the formula to calculate the crack growth rate and obtain the initial crack growth curve through calculation.

[0136] (4) Based on the above initial values, the probability distribution of the state nodes in the dynamic Bayesian network is set, and the distribution particle groups of various uncertainty parameters are generated according to the random combination method.

[0137] (5) Based on the crack length information of the current welding node obtained by deep learning system analysis, likelihood estimation is performed on each particle.

[0138] (6) Update the weights of the particles and output the updated state node parameter values ​​in the dynamic Bayesian network.

[0139] Example 2: Prediction of Remaining Life of Solder Joints

[0140] This embodiment uses a global dynamic measurement method for fatigue cracks based on multiple sensors and deep learning. A typical laboratory weld joint fatigue crack growth test is used to collect acoustic emission signals, strain, and image information at different fatigue crack lengths. The finite element method is then used to simulate the acoustic emission signals, strain, and image information at different fatigue crack lengths in weld joints.

[0141] Establish a deep learning neural network training sample set based on test and finite element simulation data; build a deep learning neural network model based on the Tensorflow program package and conduct training and verification, and establish a deep learning model for fatigue crack identification using multi-source information fusion;

[0142] Using a deep learning model, the crack length is obtained based on the acoustic emission signal, strain and image information of the crack to be tested. The crack length is then input into the railway steel bridge welding node degradation model to predict the remaining life of the weld.

[0143] In a preferred embodiment, the deep learning model includes a first branch and a second branch; the first branch is used to extract acoustic emission signal and strain signal features, and the second branch is used to extract crack length image features.

[0144] Furthermore, the first branch includes a first feature extraction module, and the second branch includes a second feature extraction module and a third feature extraction module connected in series; after the output of the second feature module interacts with the input data of the first branch, the output is respectively input into the first feature extraction module and the third feature extraction module for feature extraction;

[0145] Then, the output features of the first feature extraction module and the third feature extraction module are fused, and the crack size is obtained according to the fused features.

[0146] In an exemplary embodiment, the detailed steps refer to Figure 5 ,include:

[0147] (1) Carry out crack propagation tests on typical welded joints of railway steel bridges with various structural forms, plate thicknesses and stress modes, accurately collect crack length images during crack propagation, and measure acoustic emission signals during crack propagation and strain signals in the plastic zone at the crack tip.

[0148] (2) Establish a refined finite element model of typical welded nodes of railway steel bridges with various structural forms, plate thicknesses and stress modes, accurately simulate the initiation and propagation process of fatigue cracks, and collect crack length images, acoustic emission signals and strain signals in the plastic zone at the crack tip during the crack propagation process.

[0149] (3) The strain signal and acoustic emission signal are combined into one category as a high-fidelity signal; the crack length image is combined into one category as a low-fidelity signal.

[0150] (4) Establish a deep learning neural network model, including high-fidelity and low-fidelity neural networks. Both high-fidelity and low-fidelity neural networks are composed of two layers of fully connected neural networks. Each layer of the low-fidelity neural network contains 128 neurons, and each layer of the high-fidelity neural network contains 8 neurons.

[0151] (5) Both low-fidelity data and high-fidelity data are used as input parameters for the model to train the network. The low-fidelity data set uses the aforementioned crack length results, and the high-fidelity data set uses finite element simulation results and experimental acoustic emission signals and strain signal results; the output y of the low-fidelity neural network is L Serves as feedforward input to high-fidelity neural networks.

[0152] (6) After the low-fidelity neural network is trained, it is only necessary to train the subsequent high-fidelity neural network to achieve rapid output of the results. The output of the high-fidelity neural network y H and the output y of the low-fidelity neural network L There is a residual connection between them, and the calculation formula is as follows

[0153] y H =y L +β(tanhα1·y H_L +tanhα2·y H_NL )

[0154] Where β is the residual weight, which is set to 0.1. H_L and y H_NL are the output values ​​of the linear high-fidelity neural network NNH_L and the nonlinear high-fidelity neural network NNH_NL, respectively. α1 and α2 are two parameters that need to be trained.

[0155] (7) Using image, acoustic emission, and strain sensors to monitor the acoustic emission signals, strain signals, and image signals at fatigue cracks in welded joints of railway steel bridges, the trained multi-fidelity neural network is input to obtain the actual crack size, including crack length;

[0156] (8) The crack length is input into the degradation model of the railway steel bridge welding node, and the cyclic constitutive parameters are corrected based on the crack length to complete the prediction of the remaining fatigue life of the weld point and realize the dynamic evaluation of the health status of the weld node.

[0157] In order to realize the fatigue damage evolution of railway steel bridge welding nodes subjected to random loads under time series, this embodiment establishes a dynamic Bayesian network model for fatigue crack initiation and propagation, and uses the uncertainty factors existing in the fatigue crack initiation and propagation process as the state nodes of the model to achieve accurate evaluation of fatigue crack size and remaining fatigue life under random loads; in order to reduce the impact of uncertainty factors in the fatigue crack initiation and propagation process on the evaluation accuracy, an uncertainty parameter inference algorithm based on instrumented indentation method and particle filter inference method is established, and the key material data of fatigue-hazardous parts of welding nodes collected by instrumented indentation method is used to gradually reduce the impact of multiple uncertainty factors on the fatigue crack initiation and propagation process.

[0158] Furthermore, based on real fatigue crack data collected by multiple sensors, the uncertainty parameters during fatigue crack initiation and growth are gradually updated, enabling dynamic tracking of fatigue crack growth and accurate real-time assessment of remaining fatigue life. A high-precision identification method for measured fatigue crack data based on a deep learning neural network was also established, significantly improving the efficiency and accuracy of fatigue crack identification in railway steel bridges.

[0159] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0160] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating the residual fatigue life of welded joints based on multi-source heterogeneous information fusion, characterized in that: A sensitivity analysis of the uncertainty factors in the fatigue crack initiation and propagation process was conducted using crack length as a criterion. Based on the analysis results and the Bayesian theory, a railway steel bridge weld node degradation model was established. Collect crack length images, acoustic emission signals and strain signals; Input into the deep learning neural network to obtain the crack length; The crack length is input into the railway steel bridge weld node degradation model to obtain the residual fatigue life of the weld point.

2. The residual fatigue life assessment method according to claim 1, characterized in that: Using crack length as a criterion, a sensitivity analysis of the uncertainty factors in the fatigue crack initiation and propagation process is performed, including: When the crack length is less than 0.1 mm, the local stress-strain method is used to solve the fatigue life; When the crack length is between 0.1 mm and 0.5 mm, the IBESS method is used to solve the fatigue life; When the crack length is greater than 0.5 mm, the long crack growth prediction algorithm is used to solve the fatigue life.

3. The residual fatigue life assessment method according to claim 2, characterized in that: The local stress-strain method involves determining the nominal stress spectrum at the fatigue crack of the weld based on the measured load and weld size, combining the cyclic constitutive curve of the steel and the Neuber criterion to obtain the local stress spectrum, and then calculating the residual fatigue life of the weld based on the local stress spectrum and the cumulative damage criterion. The IBESS method includes determining a stress intensity factor spectrum based on measured loads and weld dimensions, and combining crack closure parameters and a cyclic resistance curve to obtain a plastically corrected stress intensity factor amplitude at the fatigue crack. Based on the plastically corrected stress intensity factor amplitude and the crack growth curve, the remaining fatigue life of the weld point and the predicted crack length are obtained. The long crack growth prediction algorithm includes determining the stress intensity factor spectrum based on the measured load and weld size, and obtaining the effective stress intensity factor spectrum based on the crack closure parameters; based on the effective stress intensity factor spectrum and crack growth curve, the remaining fatigue life of the weld point and the predicted crack length are obtained.

4. The residual fatigue life assessment method according to claim 3, characterized in that: Obtain the monotonic constitutive curve of fatigue crack risk area of ​​welded joint based on hardness method; Based on the monotonic constitutive curve, the initial cyclic constitutive curve is obtained by the Uniform Materials method, the initial cyclic resistance curve is obtained by the Leonetti method, and the initial crack growth curve is obtained by the iLAPS method. The obtained initial curve is used as the initial parameter value of the particle filter algorithm, and the particle weight is iteratively updated by comparing the observed crack length with the predicted crack length.

5. The residual fatigue life assessment method according to claim 4, characterized in that: High-sensitivity parameters are set as state nodes in a dynamic Bayesian network, and low-sensitivity parameters are set as fixed nodes. A degradation model of railway steel bridge welded nodes for crack initiation and propagation is established.

6. The residual fatigue life assessment method according to claim 1, characterized in that: The deep learning neural network includes a first branch and a second branch; the first branch is used to extract acoustic emission signal and strain signal features, and the second branch is used to extract crack length image features.

7. The remaining fatigue life assessment method according to claim 6, characterized in that: The first branch includes a first feature extraction module, and the second branch includes a second feature extraction module and a third feature extraction module connected in series; After the output of the second feature extraction module interacts with the input data of the first branch, it is input into the first feature extraction module and the third feature extraction module respectively for feature extraction; The output features of the first feature extraction module and the third feature extraction module are fused, and the crack size is obtained according to the fused features.

8. The remaining fatigue life assessment method according to claim 7, characterized in that: The first feature extraction module, the second feature extraction module and the third feature extraction module are all composed of a two-layer fully connected neural network.

9. The remaining fatigue life assessment method according to claim 7, characterized in that: The feature fusion expression is: and H =and L +β(tanhα1·y H_L +tanhα2·y H_NL ) Where y L is the output of the second feature extraction module, y H_L is the output of the first feature extraction module, y H_NL is the output of the third feature extraction module, α1 and α2 are parameters that need to be trained, and β is the residual weight, which is preferably 0.1.

Images