Data and physics dual-drive mechanical part low-cycle fatigue life prediction method

Abstract

The invention relates to a data and physical dual-drive mechanical part low-cycle fatigue life prediction method, which comprises the following steps: S1, carrying out finite element analysis on a mechanical part, and determining the maximum stress and strain value of a dangerous part; s2, acquiring test data of low-cycle fatigue failure, designing and carrying out a low-cycle fatigue test of the mechanical part material test piece by taking stress and strain as input, and acquiring test stress, strain and low-cycle fatigue cycle test data; s3, physical knowledge of low-cycle fatigue failure is obtained, material low-cycle fatigue physical knowledge analysis is carried out in combination with stock data, and physical knowledge of fatigue life and stress is obtained; s4, establishing a physical information neural network model, and inputting test data and physical knowledge; and S5, carrying out low-cycle fatigue life prediction on multiple dangerous parts of the mechanical part. The minimum service life of the dangerous part is used as the service life of the mechanical part. The method can significantly improve the life prediction precision.

Description

Technical Field

[0001] The present invention belongs to the technical field of aerospace engines, and in particular relates to a method for predicting the low-cycle fatigue life of mechanical parts driven by both data and physics. Background Art

[0002] In engineering applications, fatigue failure is a key issue affecting the long-term reliability of materials and structures. Traditional fatigue life prediction methods rely on empirical models and simplified physical models, which often cannot accurately describe complex failure mechanisms and material behaviors under multiple stress states. With the development of machine learning technology, neural networks have been widely used in fatigue failure analysis. These methods directly predict results by learning a large number of data samples, and in theory can handle more complex situations than traditional methods. However, purely data-driven neural networks also have obvious shortcomings. First, they are extremely dependent on a large amount of high-quality data. When data is scarce or the data quality is not high, the accuracy and reliability of the model will be greatly reduced. Second, the relationship between input and output is learned only from data, which may violate some well-known physical laws or knowledge. In addition, for scenarios outside the range of training data, existing data may be overfitted, resulting in unreasonable results. Therefore, it is necessary to combine machine learning technology and physical laws to carry out fatigue failure analysis and establish a data- and physics-driven low-cycle fatigue life prediction method for mechanical parts, thereby supporting the prediction of low-cycle fatigue life for mechanical parts.

[0003] At present, fatigue life prediction methods based on machine learning have been established: ① The existing literature "A deep learning approach for low-cycle fatigue life prediction under thermal–mechanical loading based on a novel neural network model", Engineering Fracture Mechanics, uses low-cycle fatigue data of four different materials to train a neural network and predicts the low-cycle fatigue life of various materials. However, studies have shown that the prediction accuracy of this method mainly depends on the quality and quantity of training data, and it relies too much on the quality and availability of data. ② The existing literature "A data-driven fatigue life prediction method based on parameter influence", Journal of Mechanical Engineering, the fatigue life prediction method based on BP neural network established therein can effectively predict uniaxial loading, multi-axial loading, high cycle fatigue and low cycle fatigue life, but when the quality of the training data distribution is low, the prediction accuracy will also be reduced. In terms of inventions, the existing invention of a method for predicting the low-cycle fatigue life of titanium alloy structural parts (CN115859464A) proposes a method for predicting the low-cycle fatigue life of titanium alloy structural parts based on a GA-BP neural network. The invention can predict the low-cycle fatigue life of titanium alloy structural parts, but the method is completely data-driven and does not consider physical boundary conditions. The data distribution has a great influence on the prediction accuracy. The prediction results sometimes violate the laws of physics, making it impossible to make efficient and accurate predictions and difficult to achieve engineering applications. In summary, the existing methods cannot meet the needs of high-efficiency and high-precision low-cycle fatigue life prediction in engineering. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention proposes a data and physics dual-driven low-cycle fatigue life prediction method for mechanical parts to support the low-cycle fatigue life assessment of aircraft engine mechanical parts. To achieve the above purpose, the present invention adopts the following technical solution: a data and physics dual-driven low-cycle fatigue life prediction method for mechanical parts, comprising the following steps:

[0005] Step S1: determining the stress and strain distribution of the dangerous parts, including: performing finite element analysis of mechanical parts to obtain stress and strain distribution, and determining the maximum stress and strain values ​​of the dangerous parts;

[0006] Step S2: acquiring data, including: taking stress and strain as input, designing and carrying out low-cycle fatigue tests on mechanical parts material test pieces, and acquiring test stress, strain and low-cycle fatigue cycle test data;

[0007] Step S3: obtaining physical knowledge of low-cycle fatigue failure, including: conducting physical knowledge analysis of low-cycle fatigue of materials in combination with stock data, and obtaining physical knowledge of fatigue life and stress;

[0008] Step S4: Establishing a physical information neural network model, the input includes test data and physical knowledge;

[0009] Step S5: Carry out low cycle fatigue life prediction of multiple dangerous parts of mechanical parts, and take the minimum life of dangerous parts as the life of mechanical parts.

[0010] The advantages of the present invention compared with the prior art are:

[0011] (1) This invention combines the advantages of data-driven and physical-driven methods, using not only test data but also physical knowledge of materials and mechanics. By introducing physical information, this method can capture more details of material behavior and improve the prediction accuracy of fatigue life under different working conditions.

[0012] (2) The present invention can effectively deal with complex stress states by accurately simulating complex stress-strain distribution through finite element analysis and combining physical knowledge with neural network models. This method can capture more load condition characteristics and has stronger adaptability.

[0013] (3) The present invention can continuously update the test data and physical models, continuously train and optimize the neural network model, so that it can adapt to new materials and new working conditions at any time, and maintain the accuracy and cutting-edge nature of the prediction. This allows the model to dynamically adapt to new engineering needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] FIG1 is a flow chart of a method for predicting low cycle fatigue life of mechanical parts driven by both data and physics according to the present invention;

[0015] Figure 2 is a diagram of the physical neural network architecture;

[0016] Figure 3 is a flow chart of physical neural network model training. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above-mentioned purpose, the present invention adopts the following technical scheme.

[0018] The present invention provides a data and physics dual-driven low-cycle fatigue life prediction method for mechanical parts, which supports the low-cycle fatigue life assessment of aircraft engine mechanical parts. To achieve the above purpose, the present invention adopts the following technical solutions:

[0019] First, the stress and strain distribution of dangerous parts are determined through finite element analysis, and low-cycle fatigue tests are carried out on mechanical parts material test pieces to obtain stress, strain and low-cycle fatigue cycle data; then, the physical knowledge analysis of material low-cycle fatigue is carried out in combination with the stock data to obtain the physical knowledge of fatigue life and stress; finally, a physical information neural network model is established to predict the low-cycle fatigue life of mechanical parts, including the following steps:

[0020] Step S1: determine the stress and strain distribution of dangerous parts, carry out finite element analysis of mechanical parts to obtain stress and strain distribution, and determine the maximum stress and strain position and value of dangerous parts;

[0021] Step S2: Acquire data, using the stress and strain of step S1 as input, design and carry out low-cycle fatigue tests on mechanical parts material test pieces, and acquire stress, strain and low-cycle fatigue cycle test data;

[0022] Step S3: Acquire physical knowledge, conduct physical knowledge analysis of low-cycle fatigue of materials in combination with stock data, and acquire physical knowledge of fatigue life and stress;

[0023] Step S4: Establishing a physical information neural network model, wherein the test data and physical knowledge are determined by steps S2 and S3 respectively;

[0024] Step S5: Carry out low cycle fatigue life prediction of mechanical parts, taking the minimum life of dangerous parts as the life of mechanical parts.

[0025] Furthermore, the specific steps of step S1 are:

[0026] Step S11: First, a three-dimensional model is established according to the two-dimensional design drawings of the mechanical parts;

[0027] Step S12: Set material properties in the finite element analysis software, including elastic modulus, Poisson's ratio, yield strength, etc. If the nonlinear behavior of the material is significant in low-cycle fatigue, a plasticity model, such as a bilinear or multilinear hardening model, should also be considered.

[0028] Step S13: Mesh the parts. In order to ensure the calculation accuracy, a denser mesh is used, especially in stress concentration areas (such as holes, grooves, and sharply changing geometric shapes).

[0029] Step S14: Set boundary conditions according to the actual load conditions of the mechanical parts, establish a finite element model (FEM), and apply static or dynamic loads to the finite element model according to the actual working conditions. This may include axial loads, torsional loads, bending loads, or combined loads. The loading process can simulate a single large load or multiple cyclic loads to observe the stress-strain response of the mechanical parts. If the parts are subjected to multi-axial loads (such as tension-torsion combination) in actual use, multi-axial stress analysis is required to more accurately reflect the stress state.

[0030] Step S15: Run finite element simulation to obtain stress-strain distribution. The results include key parameters such as von Mises stress, maximum principal stress, shear stress, and plastic strain. Through post-processing tools, the results are visualized and analyzed to generate stress cloud maps, deformation cloud maps, etc., to determine the concentrated areas of stress and strain in the parts. These areas are potential fatigue risk areas. Extract the stress and strain values ​​of the risk areas as input data for subsequent experiments and model analysis. If the risk area involves multiple areas, the stress and strain characteristics of each area should be recorded separately.

[0031] Compare the finite element simulation results with the actual working conditions or known test data. If there are deviations, it may be necessary to redefine the boundary conditions, optimize the meshing, or adjust the material model. Ensure that the simulation results are within a reasonable range, especially the stress and strain distribution of dangerous parts should conform to physical laws to avoid over-simplification or model distortion.

[0032] Through the finite element analysis results, the stress concentration areas and dangerous parts of mechanical parts can be identified, providing a basis for the subsequent fatigue test plan design and fatigue life prediction.

[0033] Furthermore, the specific steps of step S2 are:

[0034] Step S21: First determine the test standards and specifications that the test needs to be based on, and then formulate a test plan.

[0035] Step S22: Determine the test conditions according to the stress and strain values ​​of the dangerous parts determined in step S15. The test design should simulate the load and environment under actual working conditions as much as possible. Select suitable material specimens. The specimen shape can be a standard tensile specimen, a flat specimen, or a specially processed feature simulation specimen. The size of the specimen should meet the requirements of the testing machine and ensure uniform stress distribution. Determine the loading method and load level. Low-cycle fatigue tests usually apply cyclic loads, and the number of cycles ranges from dozens to thousands, depending on the material properties and design requirements.

[0036] Step S23: Apply cyclic load to the specimen on the testing machine. The testing machine should be equipped with high-precision strain gauges, strain gauges or displacement sensors to monitor the strain response of the specimen in real time. During the test, record the stress, strain value, and corresponding number of cycles for each loading cycle until the specimen breaks or is significantly damaged. Collect the raw data obtained in the test, including the stress-strain curve and number of cycles for each cycle. Process and organize the data to generate stress-strain curves, fatigue life curves (such as SN curves) and other key data charts. If necessary, perform data smoothing to eliminate noise and errors in the test. If there are many test samples, statistical analysis can be performed on the data under different conditions to extract the distribution characteristics of fatigue life, such as mean life, standard deviation, confidence interval, etc.

[0037] Furthermore, the specific steps of step S3 are:

[0038] Step S31: First, carry out stock data analysis to collect historical fatigue data and literature related to the current materials and working conditions, including published low-cycle fatigue test results, fatigue curves, material constitutive models, etc. Analyze the existing data and extract fatigue behaviors under stress and strain ranges similar to the current research. For example, analyze the fatigue limit and fatigue crack growth law of similar materials.

[0039] Step S32: Combine the test data and stock data, and use classical fatigue theory (such as Coffin-Manson physical model, Basquin physical model) to analyze the fatigue behavior of the material. Determine the material constants and fatigue parameters in these physical models by fitting the test data. If the material exhibits obvious nonlinear behavior, it may be necessary to adopt a nonlinear fatigue model or introduce fracture mechanics theory to more accurately describe the fatigue process. Use physical models to predict fatigue life at different stress levels, and form stress-life (SN) curves, strain-life (ε-N) curves, etc.

[0040] Step S33: Combine the test data with the results of the physical model to form a comprehensive low-cycle fatigue life knowledge base. The knowledge base should contain fatigue life information under different stress levels, strain amplitudes, loading frequencies, etc., as well as details such as crack growth rate and fatigue fracture mechanism. Analyze the relationship between stress, strain, number of cycles and fatigue life, extract key physical laws and empirical formulas, and provide physical constraints for the subsequent construction of the neural network model.

[0041] Furthermore, the specific steps of step S4 are:

[0042] Step S41: Design a suitable neural network model architecture based on data characteristics and prediction tasks. Adjust the model structure parameters, such as determining the number of network layers, number of neurons, activation function, learning rate, etc., to ensure that the model has sufficient expressive power to capture complex fatigue behaviors.

[0043] Step S42: In the design of the neural network model, physical constraints or prior knowledge are introduced. Physical knowledge is used as the input feature of the neural network model, or physical consistency constraints are introduced through regularization terms to ensure that the output of the neural network model conforms to physical laws.

[0044] Step S43: In order to incorporate physical knowledge, the training of the neural network is formulated as a constrained optimization problem.

[0045] Step S44: Customize the loss function to transform the constrained optimization problem into an unconstrained optimization problem.

[0046] Step S45: preprocess the test data obtained in step S2. This includes data cleaning (removing outliers and noise) and normalization (scaling the data to the same range) to ensure that the data is easier to learn in the neural network model. According to physical laws and fatigue theory, characteristic parameters that affect fatigue life prediction, such as maximum stress, strain amplitude, plastic strain range, etc., are extracted from data such as stress, strain, and number of cycles. The model is trained using the preprocessed test data. During the training process, the model continuously adjusts internal parameters (such as weights and biases) through the back propagation algorithm to minimize the prediction error. The cross-validation method is used to ensure that the neural network model performs well on both the training set and the validation set to avoid overfitting. During the training process, indicators such as loss function and accuracy are monitored in real time to determine the convergence of the model. If the model performance is poor, the model can be improved by increasing training data, introducing data enhancement techniques (such as adding noise, smoothing), or modifying the network architecture.

[0047] Analyze the prediction results of the model, calculate the prediction error, correlation coefficient, mean square error (MSE) and other evaluation indicators, and judge the generalization ability of the model. If the model has a large error under certain specific conditions, it may be necessary to further optimize the model or adjust the data set.

[0048] Furthermore, the specific steps of step S5 are:

[0049] Step S51: Input the test data of the dangerous parts determined in step S1 into the neural network model trained in step S4 to predict the life. For different stress levels and strain amplitudes, the model will output the corresponding fatigue life prediction values. Generally, the overall life of a part is considered to be the minimum life of its most dangerous part (the part with the maximum stress and strain).

[0050] Step S52: Compare the prediction results of the neural network model with the actual life of mechanical parts in use, and analyze the prediction accuracy and reliability of the neural network model. If there is a significant deviation between the actual life and the predicted life, the model may need to be retrained or the parameters adjusted. Through multiple applications and feedback, the model is gradually optimized to make its prediction results more accurate and stable.

[0051] Ultimately, a systematic fatigue life prediction process is formed, combining data and physical dual-driven models to provide reliable technical support for the design and maintenance of mechanical parts.

[0052] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0053] like Figure 1 As shown, a data and physics dual-driven low-cycle fatigue life prediction method for mechanical parts of the present invention comprises the following steps:

[0054] Step S1: Conduct finite element analysis of mechanical parts to obtain stress and strain distribution, and determine the maximum stress and strain locations and their values.

[0055] Step S11: Taking a turbine disk as an example, finite element analysis is performed based on UG and ABAQUS software, three-dimensional modeling of the turbine disk is performed in UG according to the two-dimensional drawings, and then the three-dimensional model is imported into ABAQUS for the next step of finite element analysis.

[0056] Step S12: Set the turbine disk material to GH720Li, set the material properties to multi-segment linear kinematic hardening in ABAQUS, and set other material properties, including elastic modulus, Poisson's ratio, yield strength, etc.

[0057] Step S13: Meshing the turbine disk. To ensure calculation accuracy, a denser mesh is used, especially in stress concentration areas (such as holes, mortises, and sharply changing geometric shapes).

[0058] Step S14: During finite element analysis, the boundary conditions of the turbine disk should be as close as possible to the load state in the service environment. The boundary conditions are set as follows:

[0059] 1) The temperature field distribution of the turbine disk is applied according to the function fitted by the heat transfer calculation results;

[0060] 2) Apply the corresponding rotational angular velocity according to the turbine disc speed;

[0061] 3) The centrifugal force generated by the turbine disk blades is applied in the form of pressure to the unit in the upper part of the tongue and groove;

[0062] 4) The axial and circumferential displacements of the turbine disk are constrained by applying UZ and UY = 0 on the mounting edge.

[0063] Step S15: Calculate the stress distribution of finite element analysis through ABAQUS. When performing force analysis on the turbine disk, it is necessary not only to pay attention to the stress and strain of dangerous parts, but also to extract the stress and strain for analysis of stress concentration areas (such as holes, mortise and tenon grooves, and sharply changing geometric shapes).

[0064] Step S2: Obtain data. Conduct low-cycle fatigue tests on mechanical parts material test pieces to obtain stress, strain and low-cycle fatigue cycle data. The specific steps are as follows:

[0065] The purpose of conducting low-cycle fatigue tests is to obtain stress-strain data and low-cycle fatigue life of mechanical parts materials, and to provide data support for the prediction of low-cycle fatigue life of mechanical parts.

[0066] Step S21: The test refers to GB / T 15248-2008 "Method for axial constant amplitude low cycle fatigue test of metallic materials", GB / T 26077-2010 "Method for axial strain control in fatigue test of metallic materials", HB 5195-1996 "Method for high temperature tensile test of metals", and other standards, and adjusts the clamping method, loading method and other contents according to the actual situation of the test piece.

[0067] Step S22: The sample is taken from the same batch of GH720Li blank discs. The sample is designed according to GJB and is forged isothermally. The heat treatment process is (1080-1110)℃×(2-4)h / oil quenching+650℃×24h / air cooling+760℃×16h / air cooling. Based on the results of the finite element analysis in step S1, the test conditions are determined as follows:

[0068] The same batch of turbine disk samples were tested at a temperature of 650°C, a stress ratio of 0.05, and loads of 800MPa, 900MPa, 1000MPa, 1100MPa, and 1200MPa, respectively. Five specimens were tested under each load condition.

[0069] Step S23: All samples were tested on the same MTS 370.10 hydraulic servo fatigue testing machine, with stress control as the control condition, triangular wave as the load form, and a loading frequency of 5 Hz. During the test, an extensometer was used to measure strain data throughout the test, and the number of cycles and test time were recorded after the test.

[0070] The test data obtained in step S2 can be used for training the physical information neural network to achieve the prediction of the low-cycle fatigue life of subsequent mechanical parts.

[0071] Step S3: Obtain physical knowledge. Combine the stock data to carry out material low cycle fatigue physical knowledge analysis and obtain physical knowledge of fatigue life and stress. The specific steps are as follows:

[0072] Step S31: First, carry out stock data analysis to collect historical fatigue data and literature related to the current materials and working conditions, including published low-cycle fatigue test results, fatigue curves, material constitutive models, etc. Analyze the existing data and extract fatigue behaviors under stress and strain ranges similar to those of the mechanical parts in the present invention.

[0073] Step S32: Combine the test data and stock data to analyze the fatigue behavior of the material. Determine the material constants and fatigue parameters in these models by fitting the test data. Use the physical model to predict the fatigue life under different stress levels and form a stress-life (SN) curve, strain-life (ε-N) curve, etc.

[0074] Step S33: Combine the test data with the results of the physical model to form a comprehensive low-cycle fatigue life knowledge base. Analyze the relationship between stress, strain, number of cycles and fatigue life, and extract key physical laws and empirical formulas. Mainly include: as the stress amplitude decreases, the variance of fatigue life increases; fatigue life shows a monotonically decreasing trend with the increase of stress amplitude; when the stress amplitude decreases to the fatigue limit, the curvature of the SN curve decreases with the decrease of stress.

[0075] Step S4: Establish a physical information neural network model, in which the test data and physical knowledge are determined by steps S2 and S3 respectively. The specific steps are:

[0076] Step S41: The physical neural network model architecture used in the present invention is as follows: Figure 2 As shown in FIG. 1 , the network includes three layers: input layer, hidden layer and output layer. The selection of the number of hidden layers is determined by the complexity of the problem. The present invention uses a network structure with one hidden layer. Unlike conventional neural networks that only learn the average value of the output from the collected distributed data, the output layer has two output neurons: the average value and the standard deviation. The loss function uses the negative log-likelihood function, that is:

[0077] (1)

[0078] In the formula, n is the number of training data, x and y are the input and output variables respectively, and θ is the set of neural network parameters. , represents the conditional probability density function of the output variable y given the input variable x and the neural network parameter θ, , which represents the cumulative distribution function of the output variable y given the input variable x and the neural network parameter θ, is the failure index, expressed as:

[0079] (2)

[0080] For the input layer, the definition of the neural network does not use an activation function. The hyperbolic tangent activation function (tanh) is used for the neurons in the hidden layer. The expression of the tanh activation function is:

[0081] (3)

[0082] Where z is the output of the corresponding neuron.

[0083] The mean of the output layer uses a linear (i.e., identity) activation function, and the expression of the linear activation function is:

[0084] (4)

[0085] Since the standard deviation is a non-negative value, the exponential linear unit (ELU) activation function is selected. The expression of the ELU activation function is:

[0086] (5)

[0087] Step S42: The network input is the stress amplitude x, and the output is the mean μ and standard deviation σ of the predicted fatigue life y. In order to ensure that the physical indicators are consistent with the basic physical knowledge, the physical knowledge determined in step S3 is expressed as physical constraints by calculating the derivative of the physical indicators with respect to the stress amplitude and applying the conditions that need to be met, i.e., physical constraints, as follows:

[0088] (1) As the stress amplitude decreases, the variance of fatigue life increases: the first-order derivative of the standard deviation is negative:

[0089] (6)

[0090] (2) Fatigue life decreases monotonically with increasing stress amplitude: the first-order derivative of the mean is negative:

[0091] (7)

[0092] (3) The curvature of the stress-life curve decreases as the stress amplitude decreases: the second-order derivative of the mean is positive:

[0093] (8)

[0094] Step S43: In order to incorporate physical knowledge, the training of the neural network is formulated as a constrained optimization problem.

[0095] (9)

[0096] Note that the above three constraints need to be met at the same time.

[0097] Step S44: The penalty function method is used to solve the optimization problem in equation (9), and the constrained optimization problem is transformed into an unconstrained optimization problem. Therefore, the composite loss function is defined as:

[0098] (10)

[0099] Where L0(θ) is the negative log-likelihood function in equation (1), and L1(θ) is the loss function corresponding to the first physical constraint, expressed as:

[0100] (11)

[0101] L2(θ) is the loss function corresponding to the second physical constraint, expressed as:

[0102] (12)

[0103] L3(θ) is the loss function corresponding to the third physical constraint, expressed as:

[0104] (13)

[0105] Penalty Factor , , is to maintain a balance between fitting the data and the physical constraints, which are enforced to be as close to zero as possible. Usually, the penalty factor , , If the value is too small, the physical constraint cannot be guaranteed. The penalty factor , , Too large a value may lead to overfitting. Therefore, by choosing an appropriate penalty factor , , value, a better fitting result can be achieved.

[0106] Step S45: In summary, the training flow chart of the physical neural network is as follows: Figure 3 As shown in the figure, the specific steps of physical neural network prediction of fatigue life are as follows:

[0107] Step S451: Divide the original data into a training set, a test set, and a validation set, and start training the neural network. , , Initially 1. When the data fitting loss converges, the neural network training stops. The physical constraint loss at the end of step S451 may not be zero.

[0108] Step S452: Determine whether the physical constraints are met. When the physical constraints are met, the whole process terminates. If not, increase the penalty factor to gradually adjust the neural network to meet the physical constraints while keeping the data fitting loss unchanged. Reduce the learning rate to prevent the updated neural network weights and biases from being too different from the results of the previous training. The initial weights and biases are reassigned to the values.

[0109] Step S453: After the parameters are updated, the neural network restarts training and monitors the physical constraint loss. When the physical constraint loss converges, the neural network training stops. Then go to step S452 to continue the calculation.

[0110] Step S5: Carry out low cycle fatigue life prediction of mechanical parts. The minimum life of dangerous parts is taken as the life of mechanical parts. The specific steps are:

[0111] Step S51: Using the test data obtained in step S2 as input, import it into the physical neural network model trained in step S4 to predict the low cycle fatigue life, and take the minimum life of all dangerous parts as the life of the mechanical parts.

[0112] Step S52: Compare the prediction results of the model with the actual life of mechanical parts in use, and analyze the prediction accuracy and reliability of the neural network model. If there is a significant deviation between the actual life and the predicted life, the neural network model may need to be retrained or the parameters adjusted. Through multiple applications and feedback, the model is gradually optimized to make its prediction results more accurate and stable.

[0113] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A data and physics-driven method for predicting the low-cycle fatigue life of mechanical parts, characterized in that: The steps include: Step S1: determining the stress and strain distribution of the dangerous parts, including: performing finite element analysis of mechanical parts to obtain stress and strain distribution, and determining the maximum stress and strain values ​​of the dangerous parts; Step S2: acquiring data, including: taking stress and strain as input, designing and carrying out low-cycle fatigue tests on mechanical parts material test pieces, and acquiring test stress, strain and low-cycle fatigue cycle test data; Step S3: obtaining physical knowledge of low-cycle fatigue failure, including: conducting physical knowledge analysis of low-cycle fatigue of materials in combination with stock data, and obtaining physical knowledge of fatigue life and stress; Step S4: Establishing a physical information neural network model, the input includes test data and physical knowledge; Step S5: Carry out low cycle fatigue life prediction of multiple dangerous parts of mechanical parts, and take the minimum life of dangerous parts as the life of mechanical parts.

2. A data and physics dual-driven low-cycle fatigue life prediction method for mechanical parts according to claim 1, characterized in that: The specific steps of step S1 are: Step S11: Building a three-dimensional model of the mechanical part according to the two-dimensional design drawing of the mechanical part; Step S12: setting the material properties of the mechanical parts in the finite element analysis software, including elastic modulus, Poisson's ratio, and yield strength; Step S13: meshing the part; using a denser mesh in stress concentration areas, which include holes, grooves, and sharply changing geometric shapes; Step S14: setting boundary conditions according to the actual load conditions of the mechanical parts, establishing a finite element model, and applying static or dynamic loads to the finite element model according to the actual working conditions; Step S15: Run finite element simulation to obtain stress-strain distribution; the results include key parameters such as von Mises stress, maximum principal stress, shear stress, and plastic strain; use post-processing tools to visualize the results and generate stress cloud maps and deformation cloud maps to determine the concentrated areas of stress and strain in mechanical parts. These areas are potential fatigue risk areas, and the stress and strain values ​​of the risk areas are extracted as input data. If the risk area involves multiple areas, the stress and strain characteristics of each area are recorded separately.

3. The method for predicting low cycle fatigue life of mechanical parts driven by data and physics according to claim 2 is characterized in that: The specific steps of step S2 are: Step S21: first determine the test standards and specifications that the test needs to be based on, and then formulate a test plan; Step S22: Determine the test conditions according to the determined stress and strain values ​​of the dangerous parts, simulate the load and environment under the actual working conditions, select the appropriate material sample, the sample shape is a standard tensile sample, a flat sample or a specially processed characteristic simulation sample, the size of the sample meets the requirements of the testing machine, and ensure that the stress distribution is uniform, determine the loading method and load level, and apply cyclic load in the low-cycle fatigue test; Step S23: applying a cyclic load to the sample on a testing machine, the testing machine is equipped with a strain gauge, a strain meter or a displacement sensor to monitor the strain response of the sample in real time. During the test, the stress and strain values ​​of each loading cycle and the corresponding number of cycles are recorded until the sample breaks or is obviously damaged. The raw data obtained in the test are collected, including the stress-strain curve and the number of cycles of each cycle, and the data are processed and sorted to generate a stress-strain curve diagram and a fatigue life curve; Statistical analysis is performed on the data under different conditions to extract the distribution characteristics of fatigue life, including mean life, standard deviation, and confidence interval.

4. The method for predicting low cycle fatigue life of mechanical parts driven by data and physics according to claim 3 is characterized in that: The specific steps of step S3 are: Step S31: First, carry out stock data analysis, collect historical fatigue data and literature related to current materials and working conditions, including low-cycle fatigue test results, fatigue curves, and material constitutive models, analyze the data, and extract the fatigue behavior of mechanical parts under the stress and strain range; Step S32: Analyze the fatigue behavior of the material by combining the test data and the stock data; By fitting the test data, the material constants and fatigue parameters in the physical model are determined, and the physical model is used to predict the fatigue life under different stress levels to form a stress-life curve, namely the SN curve, and a strain-life curve, namely the ε-N curve; Step S33: Combine the test data with the results of the physical model to form a low-cycle fatigue life knowledge base, analyze the relationship between stress, strain, number of cycles and fatigue life, and extract key physical laws; including: as the stress amplitude decreases, the variance of fatigue life increases; fatigue life shows a monotonically decreasing trend as the stress amplitude increases; When the stress amplitude decreases to the fatigue limit, the curvature of the SN curve decreases as the stress decreases.

5. The method for predicting low cycle fatigue life of mechanical parts driven by data and physics according to claim 4 is characterized in that: The specific steps of step S4 are: Step S41: Set the network architecture to include three layers: input layer, hidden layer and output layer. Set the number of hidden layers to 1. The output layer has two output neurons: mean and standard deviation. The loss function uses the negative log-likelihood function, that is: (1) In the formula, n is the number of training data, x and y are the input and output variables respectively, and θ is the set of neural network parameters. , represents the conditional probability density function of the output variable y given the input variable x and the neural network parameter θ, , which represents the cumulative distribution function of the output variable y given the input variable x and the neural network parameter θ, is the failure index, expressed as: (2) For the input layer, the definition of the neural network does not use an activation function. The hyperbolic tangent activation function tanh is used for the neurons in the hidden layer. The expression of the tanh activation function is: (3) Where z is the output of the corresponding neuron; The mean of the output layer uses a linear activation function, and the expression of the linear activation function is: (4) Select the exponential linear unit elu activation function. The expression of the elu activation function is: (5) Step S42: the network input variable x is the stress amplitude, and the output variable y is the mean μ and standard deviation σ of the predicted fatigue life; The physical knowledge determined in step S3 is expressed as a physical constraint by calculating the derivative of the physical index with respect to the stress amplitude and imposing conditions that need to be satisfied, i.e., physical constraints, as follows: 1) As the stress amplitude decreases, the variance of fatigue life increases: the first derivative of the standard deviation is negative: (6) 2) Fatigue life decreases monotonically with increasing stress amplitude: the first-order derivative of the mean is negative: (7) 3) The curvature of the stress-life curve decreases as the stress amplitude decreases: the second-order derivative of the mean is positive: (8) Step S43: In order to combine physical knowledge, the training of the neural network is expressed as a constrained optimization problem: (9) The above three constraints need to be met at the same time; Step S44: The penalty function method is used to solve the optimization problem in equation (9), and the constrained optimization problem is transformed into an unconstrained optimization problem. Therefore, the composite loss function is defined as: (10) Where L0(θ) is the negative log-likelihood function in equation (1), and L1(θ) is the loss function corresponding to the first physical constraint, expressed as: (11) L2(θ) is the loss function corresponding to the second physical constraint, expressed as: (12) L3(θ) is the loss function corresponding to the third physical constraint, expressed as: (13) in, , , is the penalty factor; Step S45: training the physical information neural network according to the above network settings for subsequent prediction of low cycle fatigue life of mechanical parts.

6. A data and physics dual-driven low-cycle fatigue life prediction method for mechanical parts according to claim 5, characterized in that: The specific steps of step S45 are: Step S451: Divide the original data into a training set, a test set, and a validation set, and start training the neural network; the penalty factor , , Initially 1, when the data fitting loss converges, the neural network training stops; Step S452: Determine whether the physical constraints are met. If the physical constraints are met, the whole process is terminated; if not, increase the penalty factor to gradually adjust the neural network to meet the physical constraints while keeping the data fitting loss unchanged, reduce the learning rate, prevent the updated neural network weights and biases from being too different from the results of the previous training, and reallocate the initial weights and biases to the values; Step S453: After the parameters are updated, the neural network restarts training and monitors the physical constraint loss. When the physical constraint loss converges, the neural network training stops and then goes to step S452 to continue the calculation.

7. A data and physics dual-driven low-cycle fatigue life prediction method for mechanical parts according to claim 6, characterized in that: The specific steps of step S5 are: Step S51: using the test data obtained in step S2 as input, importing it into the physical neural network trained in step S4 to predict the low cycle fatigue life, and taking the minimum life of all dangerous parts as the life of the mechanical parts; Step S52: Compare the prediction results of the neural network model with the actual life of mechanical parts in use, and analyze the prediction accuracy and reliability of the neural network model; if there is a significant deviation between the actual life and the predicted life, retrain the neural network model or adjust the parameters, and gradually optimize the neural network model through multiple applications and feedback.

Images