Structure fatigue life prediction method and device, equipment and storage medium

Abstract

The invention discloses a structure fatigue life prediction method, device and equipment and a storage medium. The prediction method comprises the following steps: acquiring real crack data and multiple groups of crack propagation parameters under the same load cycle index; calculating the weight according to the real crack data and the virtual crack data in each group of crack propagation parameters, and further determining the reserved crack propagation parameters; whether virtual crack data in the reserved crack propagation parameters reach critical crack data or not is judged, and if yes, the structural fatigue life is output; and otherwise, expanding the reserved crack expansion parameters by using the probability fatigue crack expansion model to obtain expanded crack expansion parameters, and repeating the steps of calculating the weight, determining the reserved expanded crack expansion parameters and judging whether the critical crack data is reached or not. According to the method, effective association between the real crack data and the probability fatigue crack propagation model is realized, and the prediction precision of the crack fatigue life is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of structural fatigue performance evaluation, and particularly relates to a structural fatigue life prediction method, device, equipment and storage medium integrating computer vision and fracture mechanics. Background Art

[0002] Traditional structural crack detection methods include X-ray method, ultrasonic method, magnetic flux leakage method, etc. The detection results of these methods are greatly affected by the subjectivity of the inspectors. With the rapid development of computer vision technology, the structural crack detection method based on computer vision can extract the image features of the surface cracks of the structure, avoiding the subjectivity and inefficiency of manual detection, and having good prospects for automation and intelligence. However, structural cracks have characteristics such as microscopicity, multiplicity and concealment, and the inhomogeneity of the structure exacerbates the uncertainty of crack initiation and propagation, resulting in obvious technical bottlenecks in the structural crack detection method based on computer vision. There is an urgent need to develop an accurate and efficient structural fatigue crack detection method based on computer vision.

[0003] The fatigue crack growth model (FCG model) has been widely used in fatigue crack growth analysis and fatigue life prediction in fields such as aerospace, automotive, and mechanical engineering. However, the fatigue crack growth behavior of structures has a high degree of randomness, resulting in a large deviation between the fatigue life prediction results based on the deterministic FCG model and the actual fatigue life. On this basis, the probabilistic fatigue crack growth model (PFCG model) was proposed to characterize the random growth behavior of cracks. Although the PFCG model considers the uncertainty of fatigue crack growth parameters through prior information, the calculated fatigue life still has a strong discreteness problem, and the evaluation results fail to meet the standards of practical applications. Therefore, due to the lack of an effective association with real crack data, there is still a large room for improvement in the fatigue life prediction accuracy of the PFCG model.

[0004] Although computer vision has obvious advantages in real-time monitoring of crack data, due to the lack of an association mechanism with the fatigue crack growth process, inspectors still cannot directly conduct real-time structural evaluation and maintenance decision-making on-site based on crack image information, which greatly limits the application potential of computer vision in structural fatigue performance evaluation. Therefore, how to fuse the measured crack image data into the PFCG model is the research focus in the field of structural fatigue performance evaluation. In fact, the application of computer vision technology in structural fatigue life prediction is extremely scarce. Summary of the Invention

[0005] The purpose of the present invention is to provide a structural fatigue life prediction method, device, equipment and storage medium to solve the problem that the PFCG model lacks an effective association with real crack data, resulting in low accuracy of crack fatigue life prediction.

[0006] The present invention solves the above technical problems through the following technical solutions: A structural fatigue life prediction method, comprising:

[0007] Obtain real crack data and multiple groups of crack growth parameters under the same load cycle number; wherein, the virtual crack data in each group of the crack growth parameters is obtained by expanding using a probabilistic fatigue crack growth model;

[0008] Calculate the weight of the corresponding virtual crack data according to the real crack data and the virtual crack data in each group of crack growth parameters under the same load cycle number;

[0009] Determine the retained crack growth parameters according to the weights of all virtual crack data;

[0010] Judge whether the virtual crack data in the retained crack growth parameters reaches the critical crack data. If so, use the corresponding load cycle number as the structural fatigue life; if not, use the probabilistic fatigue crack growth model to expand the retained crack growth parameters to obtain the expanded virtual crack data and the expanded crack growth parameters, and repeat the steps of calculating the weight of the expanded virtual crack data, determining the retained expanded crack growth parameters, and judging whether the critical crack data is reached.

[0011] Further, the process of obtaining the real crack data includes:

[0012] Obtain a real crack image, use a crack segmentation model to identify the real crack image to obtain a real crack;

[0013] Adopt a skeletonization algorithm to simplify the real crack into a crack line with a single-pixel width, and convert the crack line in the pixel coordinate system to the Cartesian coordinate system;

[0014] Fit the crack line in the Cartesian coordinate system with multiple straight line segments, and calculate the pixel size of each straight line segment;

[0015] According to the scale factor of the real crack image, convert the pixel size of each straight line segment to a physical size to obtain real crack data.

[0016] The present invention detects fatigue cracks based on computer vision, provides real-time real crack data for fatigue life prediction, realizes the effective association between fatigue life prediction and real crack data, and improves the accuracy of fatigue life prediction.

[0017] Further, the crack segmentation model is obtained by training an improved DeepLabV3+ model with a crack sample data set; the improved DeepLabV3+ model replaces the Xception convolutional layer in the original DeepLabV3+ model with a MobileNetV3, and replaces the ASPP module in the original DeepLabV3+ model with an improved ASPP module; the improved ASPP module retains the first branch and the fifth branch of the ASPP module, and redesigns the middle second branch to the fourth branch as an initialization window size layer, a first batch normalization layer, a first activation layer, a depth convolutional layer, a second batch normalization layer, and a second activation layer connected in sequence.

[0018] Using the lightweight MobileNetV3 as the backbone network, the recognition accuracy and computational efficiency are balanced, and the adaptability of the model under multi-scale targets is improved; the ASPP module in the encoder is redesigned and optimized, further reducing the model complexity.

[0019] Further, the specific construction steps of the probabilistic fatigue crack growth model include:

[0020] Obtain the stress intensity factors at different crack sizes;

[0021] Train a multi-layer perceptron using the stress intensity factors at different crack sizes to obtain a stress intensity factor prediction model;

[0022] Construct a probabilistic fatigue crack growth model according to the stress intensity factor prediction model and the Paris formula.

[0023] Further, the calculation formula for the weight of each virtual crack data is:

[0024] ;

[0025] ;

[0026] ;

[0027] where represents the standard deviation of the difference between the real crack data and each virtual crack data, represents the standard deviation function, represents the real crack data at the k-th moment, represents the i-th virtual crack data at the k-th moment, represents the likelihood function, represents the weight of the i-th virtual crack data at the k-th moment, and n represents the number of virtual crack data.

[0028] Further, determining the retained crack propagation parameters according to the weights of all virtual crack data includes:

[0029] Calculating n cumulative distribution function values according to the weights of all virtual crack data; wherein, the specific calculation formula for the m-th cumulative distribution function value is:

[0030] ;

[0031] Wherein, represents the cumulative distribution function value of the first m virtual crack data, that is, the m-th cumulative distribution function value; represents the weight of the i-th virtual crack data at the k-th moment; n represents the number of virtual crack data;

[0032] Generating n random numbers uniformly distributed in the interval [0, 1];

[0033] When the j-th random number is between the m-th cumulative distribution function value and the m + 1 cumulative distribution function value, retain the (m + 1)-th virtual crack data, and thus retain the (m + 1)-th set of crack propagation parameters.

[0034] Based on the same concept, the present invention also provides a structural fatigue life prediction device, including:

[0035] An acquisition unit for acquiring real crack data and multiple sets of crack propagation parameters under the same number of load cycles; wherein, the virtual crack data in each set of the crack propagation parameters is obtained by expanding using a probabilistic fatigue crack propagation model;

[0036] A calculation unit for calculating the weights of the corresponding virtual crack data according to the real crack data under the same number of load cycles and the virtual crack data in each set of crack propagation parameters;

[0037] A determination unit for determining the retained crack propagation parameters according to the weights of all virtual crack data;

[0038] A judgment unit for judging whether the virtual crack data in the retained crack propagation parameters reaches the critical crack data. If so, use the corresponding number of load cycles as the structural fatigue life; if not, sequentially call the acquisition unit, the calculation unit, the determination unit, and the judgment unit;

[0039] An acquisition unit is configured to acquire real crack data and multiple sets of extended crack growth parameters under the same number of load cycles; wherein, the virtual crack data in each set of extended crack growth parameters is obtained by extending the virtual crack data in the retained crack growth parameters using a probabilistic fatigue crack growth model; a calculation unit is configured to calculate the weight of the corresponding extended virtual crack data according to the real crack data under the same number of load cycles and the virtual crack data in each set of extended crack growth parameters; a determination unit is configured to determine the retained extended crack growth parameters according to the weights of all the extended virtual crack data; a judgment unit is configured to judge whether the virtual crack data in the retained extended crack growth parameters reaches the critical crack data, if so, take the corresponding number of load cycles as the structural fatigue life; if not, continue to call the acquisition unit, the calculation unit, the determination unit and the judgment unit in sequence.

[0040] Based on the same concept, the present invention further provides an electronic device, including a memory, a processor, and a computer program / instructions stored on the memory, and the processor executes the computer program / instructions to implement the structural fatigue life prediction method as described above.

[0041] Based on the same concept, the present invention further provides a computer-readable storage medium, on which a computer program / instructions is stored, and when the computer program / instructions is executed by a processor, the structural fatigue life prediction method as described above is implemented.

[0042] Compared with the prior art, the advantages of the present invention are as follows:

[0043] The present invention realizes the effective association between real crack data and a probabilistic fatigue crack growth model by acquiring real crack data and virtual crack data under the same number of load cycles, and further realizes the dynamic update of the uncertain parameters of the probabilistic fatigue crack growth model, improving the prediction accuracy of crack fatigue life. Description of the Drawings

[0044] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only one embodiment of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0045] Figure 1 is a flowchart of the structural fatigue life prediction method in an embodiment of the present invention;

[0046] Figure 2 is a partial sample schematic diagram in a crack sample dataset in an embodiment of the present invention;

[0047] Figure 3It is a schematic diagram of a crack image with a calibration plate in an embodiment of the present invention;

[0048] Figure 4 It is a schematic diagram of the structure of the original DeepLabV3+ model in an embodiment of the present invention;

[0049] Figure 5 It is a schematic diagram of the structure of MobileNetV3 in an embodiment of the present invention;

[0050] Figure 6 It is a schematic diagram of the structure of the improved ASPP module in an embodiment of the present invention;

[0051] Figure 7 It is the comparison result of the detected crack length and the actual crack length in an embodiment of the present invention;

[0052] Figure 8 It is the interaction technical flow chart of ABAQUS and FRANC3D in an embodiment of the present invention; the data in the figure represents dimensions, unit: mm;

[0053] Figure 9 It is the stress intensity factor prediction model for the mid-axis points of the crack front in an embodiment of the present invention;

[0054] Figure 10 It is the stress intensity factor prediction model for the length points of the crack front in an embodiment of the present invention;

[0055] Figure 11 It is the probability fatigue crack growth model constructed by combining the stress intensity factor prediction model and the Paris formula in an embodiment of the present invention;

[0056] Figure 12 It is the crack length update process when R = 0.1 in an embodiment of the present invention;

[0057] Figure 13 It is the crack length update process when R = 0.2 in an embodiment of the present invention;

[0058] Figure 14 It is the crack length update process when R = 0.3 in an embodiment of the present invention;

[0059] Figure 15 It is the fatigue life prediction result when R = 0.1 in an embodiment of the present invention;

[0060] Figure 16 It is the fatigue life prediction result when R = 0.2 in an embodiment of the present invention;

[0061] Figure 17 It is the fatigue life prediction result when R = 0.3 in an embodiment of the present invention. Detailed implementation manners

[0062] The following clearly and completely describes the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts belong to the scope of protection of the present invention.

[0063] The following specifically describes the technical solutions of the present invention with specific embodiments. These several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0064] Embodiment 1

[0065] As Figure 1 shown, the structural fatigue life prediction method provided by the present invention includes the following steps:

[0066] Step S1: Obtain real crack data and multiple groups of crack growth parameters under the same number of load cycles.

[0067] In order to effectively associate the real crack data with the fatigue crack growth process, the present invention is based on computer vision and obtains the real crack data according to the real crack image. In the specific implementation manner of the present invention, the process of obtaining the real crack data includes:

[0068] Step A1: Obtain a real crack image, use a crack segmentation model to identify the real crack image, and obtain a real crack.

[0069] The real crack image is obtained by photographing the crack on the structure by an image acquisition system, and at the same time, the number of load cycles corresponding to the real crack image can be known through a fatigue testing machine or other load cycle recording methods. The structure in this embodiment is a steel structure, and the crack segmentation model is obtained by training an improved DeepLabV3+ model with a crack sample data set. The specific training process includes:

[0070] Step A1.1: Construct a crack sample data set.

[0071] The crack sample data set includes multiple samples, each sample includes an input quantity and an output quantity. The input quantity is a scene image containing a crack (i.e., a crack image), and the output quantity is a segmentation mask used to identify whether each pixel in the input image belongs to the crack category, as Figure 2 shown. All images are uniformly processed into a fixed resolution of 512×512 and divided into a training set and a validation set according to a ratio of 9:1. The training set includes 3919 crack images, and the validation set includes 436 crack images.

[0072] To avoid the inability to convert the pixel size of cracks into physical size (i.e., real-world size) due to the lack of a reference measurement benchmark object in the sample, the present invention conducts fatigue crack growth analysis on compact tension specimens (i.e., CT specimens), and collects a large number of images with calibration plates, such as Figure 3 shown.

[0073] Specifically, the CT specimens are made of Q345qd steel, with an elastic modulus of 2.10×105 MPa and a Poisson's ratio of 0.3. The experimental equipment mainly includes a fatigue testing machine and an image acquisition system. The CT specimens are connected to the clamping end of the brake through a prefabricated fixture and a bolt pin, and a sinusoidal fatigue cyclic load of 7 Hz is applied in a force control manner. The stress ratio (R) is controlled by fixing the upper limit of the fatigue load and adjusting the lower limit of the fatigue load, and the maximum load is taken as 5 kN. The image acquisition system mainly consists of 1 CCD camera, 1 LED light source and interactive software. To ensure that the camera can clearly collect crack images, the camera is set at a position 0.5 m away from the specimen surface, with a resolution of 4096×3000 pixels and a frame rate of 40 fps. To accelerate the formation of fatigue cracks, a wire cutting technique is used to prefabricate an initial defect of 2 mm on each CT specimen.

[0074] Step A1.2: Construct an improved DeepLabV3+ model.

[0075] Although the original DeepLabV3+ model can be directly used for structural crack detection, the original DeepLabV3+ model is difficult to meet the requirements of lightweight detection. Therefore, the present invention improves the original DeepLabV3+ model. The structure of the original DeepLabV3+ model is as Figure 4 shown. On the one hand, a lightweight MobileNetV3 is used as the backbone network to replace the original Xception convolutional layer to balance the recognition accuracy and computational efficiency, thereby improving the adaptability of the model under multi-scale targets. The structure of MobileNetV3 is as Figure 5 shown.

[0076] On the other hand, the ASPP module (Atrous Spatial Pyramid Pooling module) of the original DeepLabV3+ model is redesigned, as Figure 6 shown. Specifically, the first branch and the fifth branch of the ASPP module are retained, and at the same time, the middle second branch to the fourth branch are redesigned as an initial window size layer, a first batch normalization layer, a first activation layer, a depth convolutional layer, a second batch normalization layer and a second activation layer connected in sequence. The redesigned ASPP module only focuses on the local correlation of the current window and its neighboring windows, without considering the global relationship of the entire image, effectively reducing the computational overhead from global to local, and enhancing the scalability of the model.

[0077] Step A1.3: Train and validate the improved DeepLabV3+ model using the crack sample dataset constructed in Step A1.1 to obtain a crack segmentation model.

[0078] Train the improved DeepLabV3+ model using the training set and validate the improved DeepLabV3+ model using the validation set. During the training process, the hyperparameters are set as follows: The optimizer is selected as Adam, and the initial learning rate is 5e -4 , to accelerate convergence and maintain training stability. The momentum factor is set to 0.9 to further enhance convergence and reduce oscillations during training. The Adam optimizer has the characteristic of adaptive learning rate and can dynamically adjust the learning rate according to the training process. To prevent overfitting, the weight decay coefficient is set to 0, the batch size is 8, and the number of training epochs is 100. And a TXT document is generated to record and observe the loss changes of the model on the training set and the validation set. The loss function uses cross-entropy loss, which is widely used in the crack segmentation task to optimize the accuracy of crack recognition.

[0079] Step A1.4: Evaluate the performance of the crack segmentation model.

[0080] In this embodiment, indicators such as Intersection over Union (IoU), Recall, and Precision are used to evaluate the crack recognition accuracy of the original DeepLabV3+ model and the improved DeepLabV3+ model. IoU represents the ratio of the intersection to the union of the crack prediction value and the ground truth, which intuitively reflects the crack segmentation effect. Recall measures the proportion of the number of correctly identified positive samples to all positive samples and is used to evaluate the recognition coverage ability of the model for positive samples. Precision is the proportion of actual positive samples among the predicted positive samples, which reflects the accuracy of the model's prediction results. In addition, the number of parameters, FLOPs, and training time are key indicators for evaluating the lightweight degree of the model and can measure the computational complexity and running efficiency of the model.

[0081] This embodiment is based on the PyTorch deep learning framework. In the GPU environment, the original DeepLabV3+ model and the improved DeepLabV3+ model are trained and verified using the crack sample dataset constructed in step A1.1. Then, performance quantitative evaluation is carried out on the verified original DeepLabV3+ model and improved DeepLabV3+ model using evaluation metrics. The evaluation results are shown in Table 1, where mIoU, mRecall, and mPrecision represent the average values of the intersection over union, recall rate, and precision rate, respectively. As can be seen from Table 1, the improved DeepLabV3+ model has a significant improvement in crack lightweight recognition compared to the DeepLabV3+ model, with the number of parameters, FLOPs, and training time reduced by 93.18%, 73.17%, and 35.55% respectively. In addition, while significantly reducing the computational resource requirements, the improved DeepLabV3+ model can still maintain a high crack recognition accuracy, where mIoU, mRecall, and mPrecision reach 88.57%, 94.00%, and 93.15% respectively.

[0082] Table 1 Structural crack recognition results of different models

[0083] Step A2: Use a skeletonization algorithm to simplify the real cracks obtained in step A1 into crack lines with a single-pixel width, and convert the crack lines in the pixel coordinate system to the Cartesian coordinate system.

[0084] To calculate real crack data (such as real crack length, real crack angle, etc.), this embodiment uses a skeletonization algorithm to refine the real cracks identified by the crack segmentation model into crack lines with a single-pixel width, and removes small-area noise regions through connected component analysis.

[0085] Step A3: Fit the crack lines in the Cartesian coordinate system with multiple straight line segments, and calculate the pixel size of each straight line segment.

[0086] Since the crack lines with a single-pixel width in the Cartesian coordinate system are extremely irregularly shaped lines, multiple straight line segments are used to fit the crack lines. During the fitting process, the Nelder-Mead optimization algorithm, gradient descent method, or least squares method is used to minimize the error between the skeleton points and the straight line segments to ensure the fitting accuracy. After the fitting is completed, the pixel size information of each straight line segment is calculated.

[0087] The calculation formula for the pixel length of each straight line segment is:

[0088] (1)

[0089] The calculation formula for the pixel angle of each straight-line segment is as follows:

[0090] (2)

[0091] In the formula, L represents the pixel length of the current straight-line segment, θ represents the pixel angle between the current straight-line segment and the previous straight-line segment, represents the starting point of the previous straight-line segment, represents the ending point of the previous straight-line segment and the starting point of the current straight-line segment, represents the ending point of the current straight-line segment.

[0092] Step A4: According to the scale factor of the real crack image, convert the pixel size of each straight-line segment into a physical size to obtain real crack data.

[0093] The scale factor of the real crack image is calculated by using the calibration plate corner point detection method. By means of the scale factor, the pixel size is converted into a physical size, realizing the conversion of the pixel size of the crack to the size in the real world. The specific calculation formula for the scale factor is:

[0094] (3)

[0095] Among them, A represents the scale factor of the real crack image, represents the physical length of the calibration plate, represents the pixel length of the calibration plate. Therefore, the pixel length of the straight-line segment can be converted into a physical length through the scale factor. The real crack data in this embodiment includes the real crack physical length, that is, the real crack length.

[0096] In order to verify the effectiveness of crack data acquisition, when collecting the crack images in the crack sample dataset, a scale is used for calibration, as Figure 7 shown. It can be seen from Figure 7 that the detected crack length is highly consistent with the actual crack length, and the overall recognition error is controlled within ±0.5 mm, with an average error of 0.31 mm. It shows the reliability and practicability of the crack data acquisition method of the present invention in engineering applications, and can provide real-time and accurate real crack lengths for the digital twin (DT) driving framework.

[0097] The present invention uses an improved DeepLabV3+ model for structural crack recognition, and calculates crack data based on the recognized cracks, which can provide real-time and accurate structural real crack data for the digital twin driving framework.

[0098] The fatigue crack growth (FCG) analysis using the ABAQUS and FRANC3D interaction technology can obtain virtual crack data. However, the time for executing a single FCG analysis cycle exceeds 70 minutes, resulting in an extremely low generation rate of virtual crack data and making it difficult to meet the real-time update requirements of uncertain FCG parameters (i.e., crack growth parameters). To solve this technical problem, the present invention constructs a probabilistic fatigue crack growth model (PFCG model) and uses the PFCG model to perform crack growth on the input FCG parameters to obtain virtual crack data.

[0099] In a specific embodiment of the present invention, the specific construction steps of the probabilistic fatigue crack growth model include:

[0100] Step B1: Obtain the stress intensity factors at different crack sizes.

[0101] Based on the CT specimen in step A1.1, by adjusting the inserted crack size, the corresponding stress intensity factor SIF at the crack front is calculated using the ABAQUS and FRANC3D interaction technology based on the M integral method, and thus the stress intensity factors at different crack sizes can be obtained, as Figure 8 shown. The crack sizes in this embodiment include a and h, where a represents the length of the crack mid-axis and h represents the crack length, Figure 8 and c in

[0102] represents the length of the semi-major axis of the crack.

[0103] Construct a stress intensity factor data set according to the stress intensity factors at different crack sizes. Each sample in the stress intensity factor data set includes an input quantity and an output quantity. The input quantity is the crack sizes a and h, and the output quantity is the corresponding stress intensity factor SIF. The stress intensity factor data set in this embodiment includes 3500 samples and is used to train a multi-layer perceptron (i.e., MLP).

[0104] In this embodiment, a machine learning method is used to establish a high-precision stress intensity factor prediction model, that is, the SIF prediction model, to capture the non-linear relationship between the crack sizes a and h and the SIF. 1500 input quantities are used to test the stress intensity factor prediction model, and the test results are as Figure 9 and Figure 10 shown, where represents the stress intensity factor at the mid-axis point of the crack front, represents the stress intensity factor at the length point of the crack front. As can be seen from Figure 9 and Figure 10 , the stress intensity factor prediction model can quickly and accurately output the corresponding SIF according to the given crack size.

[0105] Step B3: Construct a probabilistic fatigue crack growth model based on the stress intensity factor prediction model and the Paris formula.

[0106] Based on the prior information, establish a probability distribution model for the initial crack growth parameters (a, h, C, m) of the probabilistic fatigue crack growth model. Subsequently, sample the probability distribution model to generate groups of initial crack growth parameters. In this embodiment, the initial crack growth parameters include the crack size a and h, and the material-related parameters C and m. Assume that the material-related parameter C follows a normal distribution (3.3×10 -13 , 1.6×10 -13 ), and the material-related parameter m follows a uniform distribution (2.80, 3.10). Obtain the material-related parameters C and m according to the prior information. By simulating the crack growth process of each sampled sample through the stress intensity factor prediction model and the Paris formula, virtual crack data at any number of load cycles can be output, that is, the crack sizes a and h, as Figure 11 shown, where represents the stress intensity factor range threshold value, , and respectively represent the crack mid-axis length, crack length, and number of load cycles at the current moment.

[0107] In the PFCG model, the Paris formula is used to describe the relationship between the stress intensity factor range and the crack growth rate. The specific formula is:

[0108] , (4)

[0109] where N represents the number of load cycles, ΔK represents the stress intensity factor range (directly obtained from the SIF prediction model), and respectively represent the stress intensity factor ranges at the crack mid-axis point and crack length point varying with the number of load cycles.

[0110] The crack growth increment at a given number of load cycles can be expressed according to the integral method as:

[0111] (5)

[0112] (6)

[0113] where represents the increment in the crack mid-axis direction, represents the increment in the crack length, represents the increment in the number of load cycles. The machine learning-assisted PFCG model uses the crack length Reach the critical crack length As the termination condition for fatigue crack growth analysis. It should be noted that on a 20-core ((Intel i7-14700KF) workstation, the time for a single fatigue crack growth analysis cycle is shortened to within 1 second. Compared with the 70 minutes required by the pure ABAQUS and FRANC3D interaction technology, the calculation efficiency is significantly improved, and the generation of crack data is accelerated through the PFCG model. The crack data generated by the PFCG model is defined as virtual crack data. For example, the length of the crack centerline generated by the PFCG model is defined as the virtual crack centerline length, both represented by the character a; the crack length generated by the PFCG model is defined as the virtual crack length, both represented by the character h.

[0114] Since the real crack centerline length cannot be obtained through steps A1 to A4, therefore, both the real crack data and the virtual crack data in this embodiment refer to the crack length, that is, the real crack length and the virtual crack length.

[0115] Step S2: Calculate the weight of the corresponding virtual crack data according to the real crack data and the virtual crack data in each group of crack growth parameters under the same load cycle number.

[0116] The present invention uses a dynamic Bayesian network DBN to establish a synchronous connection between the real crack data and the crack growth parameters under the same load cycle number, and uses the particle filter algorithm as the inference method of the dynamic Bayesian network DBN.

[0117] Sample the probability distribution model of the initial crack growth parameters of the PFCG model to generate 100 groups of initial crack growth parameters (a, h, C, m). Then the particle swarm includes 100 particles, each particle corresponds to a group of initial crack growth parameters (a, h, C, m) and each particle has the same weight. According to the PFCG model, establish the state equations for fatigue crack growth analysis (Equations (7) and (8)):

[0118] (7)

[0119] (8)

[0120] Where, represents the crack centerline length at the current moment or the k-th moment (i.e., the virtual crack centerline length), represents the crack centerline length at the previous moment or the k-1-th moment, represents the state noise, represents the crack length at the current moment or the k-th moment (i.e., the virtual crack length), represents the crack length at the previous moment or the k-1-th moment, represents the true crack length at the current moment (obtained according to steps A1 to A4), represents the weight of the crack length.

[0121] The true crack length is compared with the calculated value of the state equation (i.e., the output value of the PFCG model, i.e., the virtual crack length ), the error between each virtual crack length and the true crack length is calculated, and the likelihood function value is calculated through Gaussian distribution to update the weights of each particle at the current moment:

[0122] (9)

[0123] (10)

[0124] (11)

[0125] where represents the true crack length and the standard deviation of the difference between each virtual crack length , represents the standard deviation function, represents the likelihood function, represents the weight of the i-th virtual crack length or the i-th particle at the k-th moment, n represents the number of virtual crack lengths or particles, and n = 100 in this embodiment.

[0126] Specifically, based on the fatigue crack growth analysis experiment in step A1.1, the crack images are collected by the image acquisition system at 5000 load cycles, and the acquisition period is appropriately shortened only when the crack is close to the critical length. Considering that when the true crack length reaches about 14 mm during the experiment, the CT specimen will break rapidly in a very short time. Therefore, 14 mm is determined as the critical crack length. Subsequently, the true crack length is extracted from the crack images. Correspondingly, the PFCG model also calculates the virtual crack length at 5000 load cycles. In this embodiment, 3 groups of fatigue crack growth analysis experiments are carried out under stress ratios R = 0.1, 0.2, and 0.3 respectively. The virtual crack lengths output by the PFCG model are compared with the true crack lengths at intervals of 20000, 30000, and 40000 load cycles respectively, and then the matching degree between the virtual crack length and the true crack length is calculated according to formulas (9) to (11), so as to determine the weights of each particle or each virtual crack length.

[0127] Step S3: Determine the retained crack growth parameters according to the weights of all virtual crack data.

[0128] In a specific embodiment of the present invention, the virtual crack data refers to the virtual crack length, and the crack growth parameters to be retained are determined according to the weights of all virtual crack lengths (i.e., particles), including:

[0129] Step S3.1: Calculate n cumulative distribution function values according to the weights of all virtual crack lengths; wherein, the specific calculation formula for the m-th cumulative distribution function value is:

[0130] (12)

[0131] wherein, represents the cumulative distribution function value of the first m virtual crack lengths, that is, the m-th cumulative distribution function value; represents the weight of the i-th virtual crack length or the i-th particle at the k-th moment; n represents the number of virtual crack lengths or particles. When m takes 1, the first cumulative distribution function value is obtained; when m takes 2, the second cumulative distribution function value is obtained; and so on. When m takes n, the n-th cumulative distribution function value is obtained.

[0132] Step S3.2: Generate n random numbers uniformly distributed in the interval [0, 1].

[0133] Step S3.3: When the j-th random number is between the m-th cumulative distribution function value and the (m + 1)-th cumulative distribution function value, retain the (m + 1)-th virtual crack length, and thus retain the (m + 1)-th set of crack growth parameters.

[0134] The crack growth parameters to be retained are determined according to the cumulative distribution function values and random numbers, and the number of retained crack growth parameters is still n. The greater the weight of the virtual crack length or particle, the greater the probability of being selected; conversely, the smaller the weight of the virtual crack length or particle, the smaller the probability of being selected, or even not being selected at all.

[0135] In this embodiment, the update processes of the material-related parameters C and m in 3 groups of fatigue crack growth analysis experiments are shown in Table 2, and the update process of the crack length h is as Figures 12 to 14 shown. The results show that by continuously integrating the measured crack length (i.e., the real crack length), the crack growth parameters are dynamically updated, and at the same time, the uncertainty of the crack growth parameters is reduced.

[0136] Table 2 Update processes of material-related parameters C and m

[0137] Step S4: Determine whether the virtual crack data in the reserved crack propagation parameters reaches the critical crack data. If so, use the corresponding number of load cycles as the structural fatigue life. If not, use the probabilistic fatigue crack propagation model to propagate the reserved crack propagation parameters to obtain the extended virtual crack data and the extended crack propagation parameters, and repeat steps S2 to S4.

[0138] Use the average value of all reserved virtual crack lengths (i.e., the virtual crack lengths in the reserved crack propagation parameters) as the current crack length h k , and determine this current crack length h k whether it reaches the critical crack length h f . If so, use the corresponding number of load cycles N as the structural fatigue life. If not, input the reserved crack propagation parameters into the PFCG model, perform the FCG analysis at the (k + 1)-th moment to obtain the extended virtual crack length and the extended crack propagation parameters, and then repeat steps S2 to S4 until the critical crack length is reached, and output the structural fatigue life.

[0139] In this embodiment, 3 groups of fatigue crack propagation analysis experiments each performed 3 times of DBN reasoning (that is, each group of experiments performed 3 times of comparison between the real crack length (i.e., the experimental value) and the virtual crack length output by the PFCG model), and the predicted structural fatigue lives are as Figures 15 to 17 shown. The results show that DBN reasoning establishes a synchronous connection between the real crack data and the PFCG model. By continuously fusing the real crack data, the uncertainty of the crack propagation parameters of the PFCG model is effectively reduced, and the prediction accuracy of the fatigue life is significantly improved, with an average error of 6.94%.

[0140] The present invention integrates computer vision and fracture mechanics theory to propose a digital twin-driven framework for real-time prediction of structural fatigue life. Based on the improved DeepLabV3+ model, it realizes real-time detection of structural cracks, providing real-time and accurate real crack data for the digital twin-driven framework; establishes a machine learning-based PFCG model to accelerate the generation of virtual crack data; on this basis, uses the dynamic Bayesian network (DBN) to establish a synchronous connection between virtual and real crack data in the digital space, and updates the uncertain parameters of the PFCG model in real time, thereby improving the prediction accuracy of the steel structure fatigue life, and reducing the average fatigue life prediction error from 52.33% to 6.94%. The effectiveness of the proposed digital twin-driven framework is verified through experimental research.

[0141] Embodiment 2

[0142] The structural fatigue life prediction device provided by the present invention includes an acquisition unit, a calculation unit, a determination unit, and a judgment unit.

[0143] An acquisition unit for acquiring real crack data and multiple groups of crack growth parameters under the same number of load cycles; wherein, the virtual crack data in each group of the crack growth parameters is obtained by expanding using a probabilistic fatigue crack growth model. The real crack data and virtual crack data in this embodiment respectively refer to the real crack length and the virtual crack length.

[0144] A calculation unit for calculating the weight of the corresponding virtual crack data according to the real crack data under the same number of load cycles and the virtual crack data in each group of crack growth parameters, referring to formulas (9) to (11) in Embodiment 1.

[0145] A determination unit for determining the retained crack growth parameters according to the weights of all virtual crack data.

[0146] A judgment unit for judging whether the virtual crack data in the retained crack growth parameters reaches the critical crack data. If so, the corresponding number of load cycles is used as the structural fatigue life; if not, the acquisition unit, the calculation unit, the determination unit, and the judgment unit are called in sequence. The acquisition unit is used to acquire real crack data and multiple groups of expanded crack growth parameters under the same number of load cycles; wherein, the virtual crack data in each group of the expanded crack growth parameters is obtained by expanding the virtual crack data in the retained crack growth parameters using a probabilistic fatigue crack growth model; the calculation unit is used to calculate the weight of the corresponding expanded virtual crack data according to the real crack data under the same number of load cycles and the virtual crack data in each group of the expanded crack growth parameters; the determination unit is used to determine the retained expanded crack growth parameters according to the weights of all the expanded virtual crack data; the judgment unit is used to judge whether the virtual crack data in the retained expanded crack growth parameters reaches the critical crack data. If so, the corresponding number of load cycles is used as the structural fatigue life; if not, continue to call the acquisition unit, the calculation unit, the determination unit, and the judgment unit in sequence.

[0147] In some specific embodiments of the present invention, the structural fatigue life prediction device can incorporate the features of the structural fatigue life prediction method in Embodiment 1 of the present invention, and vice versa.

[0148] Embodiment 3

[0149] The embodiment of the present invention also provides an electronic device, which includes: a memory, a processor, and a computer program / instructions stored on the memory, and the processor executes the computer program / instructions to implement the structural fatigue life prediction method in Embodiment 1 of the present invention.

[0150] Although not shown, the electronic device includes a processor, which can perform various appropriate operations and processes according to programs and / or data stored in a read-only memory (ROM) or programs and / or data loaded from a storage section into a random access memory (RAM). The processor can be a multi-core processor or can include multiple processors. In some embodiments, the processor can include a general-purpose main processor and one or more special coprocessors, such as a central processing unit, a graphics processing unit (GPU), a neural network processor (NPU), a digital signal processor (DSP), and so on. In the RAM, various programs and data required for device operation are also stored. The processor, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0151] The above-mentioned processor and memory are jointly used to execute the programs / instructions stored in the memory. When the programs / instructions are executed by a computer, they can implement the methods, steps, or functions described in the above embodiments.

[0152] Although not shown, an embodiment of the present invention also provides a computer-readable storage medium, on which computer programs / instructions are stored. When the computer programs / instructions are executed by a processor, the structural fatigue life prediction method in the first embodiment of the present invention is implemented.

[0153] The readable storage medium includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassette tapes, disk storage, or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media do not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0154] The specific embodiments disclosed above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or variations within the technical scope disclosed by the present invention and should be covered by the protection scope of the present invention.

Claims

1. A method for predicting the structural fatigue life, characterized in that The prediction method includes: Obtain the true crack data and multiple groups of crack growth parameters under the same number of load cycles; among them, the virtual crack data in each group of the crack growth parameters is obtained by expanding using a probabilistic fatigue crack growth model; Calculate the weight of the corresponding virtual crack data according to the true crack data and the virtual crack data in each group of crack growth parameters under the same number of load cycles; Determine the retained crack growth parameters according to the weights of all virtual crack data; Judge whether the virtual crack data in the retained crack growth parameters reaches the critical crack data. If so, use the corresponding number of load cycles as the structural fatigue life; if not, use the probabilistic fatigue crack growth model to expand the retained crack growth parameters to obtain the expanded virtual crack data and the expanded crack growth parameters, and repeat the steps of calculating the weights of the expanded virtual crack data, determining the retained expanded crack growth parameters, and judging whether the critical crack data is reached.

2. The structural fatigue life prediction method according to claim 1, characterized in that, The process of obtaining the true crack data includes: Obtain a true crack image, and use a crack segmentation model to identify the true crack image to obtain the true crack; Adopt a skeletonization algorithm to simplify the true crack into a crack line with a single-pixel width, and convert the crack line in the pixel coordinate system to the Cartesian coordinate system; Fit the crack line in the Cartesian coordinate system with multiple straight line segments, and calculate the pixel size of each straight line segment; According to the scale factor of the true crack image, convert the pixel size of each straight line segment to a physical size to obtain the true crack data.

3. The structural fatigue life prediction method according to claim 2, characterized in that The crack segmentation model is obtained by training an improved DeepLabV3+ model with a crack sample data set; the improved DeepLabV3+ model replaces the Xception convolutional layer in the original DeepLabV3+ model with a MobileNetV3, and replaces the ASPP module in the original DeepLabV3+ model with an improved ASPP module; the improved ASPP module retains the first branch and the fifth branch of the ASPP module, and redesigns the middle second branch to the fourth branch as an initialization window size layer, a first batch normalization layer, a first activation layer, a depth convolutional layer, a second batch normalization layer, and a second activation layer connected in sequence.

4. The structural fatigue life prediction method according to claim 1, wherein The specific construction steps of the probabilistic fatigue crack growth model include: Obtain the stress intensity factor at different crack sizes; Train a multi-layer perceptron using the stress intensity factor at different crack sizes to obtain a stress intensity factor prediction model; Construct a probabilistic fatigue crack growth model according to the stress intensity factor prediction model and the Paris formula.

5. The structural fatigue life prediction method according to claim 4, characterized in that, Adopt the ABAQUS and FRANC3D interaction technology to obtain the stress intensity factor at different crack sizes.

6. The structural fatigue life prediction method according to claim 1, characterized in that The calculation formula for the weight of each virtual crack data is: ； ； ； Among them, represents the standard deviation of the difference between the real crack data and each virtual crack data, represents the standard deviation function, represents the real crack data at the k-th moment, represents the i-th virtual crack data at the k-th moment, represents the likelihood function, represents the weight of the i-th virtual crack data at the k-th moment, and n represents the number of virtual crack data.

7. The structural fatigue life prediction method according to any one of claims 1 to 6, characterized in that, The determination of the retained crack growth parameters according to the weights of all virtual crack data includes: Calculate n cumulative distribution function values according to the weights of all virtual crack data; among them, the specific calculation formula for the m-th cumulative distribution function value is: ； Among them, represents the cumulative distribution function value of the first m virtual crack data, that is, the m-th cumulative distribution function value; represents the weight of the i-th virtual crack data at the k-th moment; n represents the number of virtual crack data; Generate n random numbers uniformly distributed in the interval [0, 1]; When the j-th random number is between the m-th cumulative distribution function value and the (m + 1)-th cumulative distribution function value, retain the (m + 1)-th virtual crack data, and thus retain the (m + 1)-th set of crack propagation parameters.

8. A structural fatigue life prediction device, characterized in that, The prediction device includes: An acquisition unit for acquiring real crack data and multiple sets of crack propagation parameters under the same number of load cycles; wherein, the virtual crack data in each set of the crack propagation parameters is obtained by expanding using a probabilistic fatigue crack propagation model; A calculation unit for calculating the weight of the corresponding virtual crack data according to the real crack data under the same number of load cycles and the virtual crack data in each set of crack propagation parameters; A determination unit for determining the retained crack propagation parameters according to the weights of all virtual crack data; A judgment unit for judging whether the virtual crack data in the retained crack propagation parameters reaches the critical crack data. If so, use the corresponding number of load cycles as the structural fatigue life; if not, sequentially call the acquisition unit, the calculation unit, the determination unit, and the judgment unit; An acquisition unit for acquiring real crack data and multiple sets of expanded crack propagation parameters under the same number of load cycles; wherein, the virtual crack data in each set of the expanded crack propagation parameters is obtained by expanding the virtual crack data in the retained crack propagation parameters using a probabilistic fatigue crack propagation model; a calculation unit for calculating the weight of the corresponding expanded virtual crack data according to the real crack data under the same number of load cycles and the virtual crack data in each set of the expanded crack propagation parameters; a determination unit for determining the retained expanded crack propagation parameters according to the weights of all the expanded virtual crack data; a judgment unit for judging whether the virtual crack data in the retained expanded crack propagation parameters reaches the critical crack data. If so, use the corresponding number of load cycles as the structural fatigue life; if not, continue to sequentially call the acquisition unit, the calculation unit, the determination unit, and the judgment unit.

9. An electronic device, comprising a memory, a processor, and computer programs / instructions stored on the memory, characterized in that, The processor executes the computer program / instructions to implement the structural fatigue life prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer programs / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, the structural fatigue life prediction method according to any one of claims 1 to 7 is implemented.

Images