CN112052615A - Micro-motion fatigue performance prediction method based on artificial neural network - Google Patents
Micro-motion fatigue performance prediction method based on artificial neural network Download PDFInfo
- Publication number
- CN112052615A CN112052615A CN202010929127.7A CN202010929127A CN112052615A CN 112052615 A CN112052615 A CN 112052615A CN 202010929127 A CN202010929127 A CN 202010929127A CN 112052615 A CN112052615 A CN 112052615A
- Authority
- CN
- China
- Prior art keywords
- neural network
- artificial neural
- fatigue
- fretting
- fretting fatigue
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000005457 optimization Methods 0.000 claims abstract description 28
- 238000002474 experimental method Methods 0.000 claims abstract description 27
- 238000004364 calculation method Methods 0.000 claims abstract description 18
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 10
- 210000002569 neuron Anatomy 0.000 claims description 69
- 230000000977 initiatory effect Effects 0.000 claims description 48
- 238000012549 training Methods 0.000 claims description 22
- 238000007635 classification algorithm Methods 0.000 claims description 13
- 238000012795 verification Methods 0.000 claims description 12
- 238000009661 fatigue test Methods 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 description 6
- 230000007547 defect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medical Informatics (AREA)
- Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)
Abstract
The invention discloses a micromotion fatigue performance prediction method based on an artificial neural network, which comprises the steps of obtaining corresponding data parameters through a series of experiments, constructing a micromotion fatigue numerical model and the artificial neural network, optimizing the artificial neural network by using a back propagation algorithm according to the error of a prediction result of the artificial neural network and a calculation result of the micromotion fatigue numerical value, so that the overall optimization is achieved, and finally, the accurate prediction of the micromotion fatigue performance is completed; according to the invention, the artificial neural network with low cost and global optimization is obtained by improving the existing artificial neural network, and finally the fretting fatigue performance is predicted according to the improved artificial neural network, so that the problems of high cost and incapability of achieving global optimization when the fretting fatigue performance is predicted by adopting the existing artificial neural network are solved, the experiment and numerical calculation cost is reduced, and the prediction precision is improved.
Description
The technical field is as follows:
the invention relates to the technical field of artificial intelligence, in particular to a micromotion fatigue performance prediction method based on an artificial neural network.
Background art:
fretting refers to the relative tangential motion of very small amplitude between the contact surfaces, nominally stationary, under alternating loads such as mechanical vibrations, fatigue loads, etc. Although the displacement amplitude of the micromotion is not strictly defined, it is generally considered to be within 100 μm. The forms of fretting include fretting fatigue, fretting wear, fretting corrosion. Fretting fatigue refers to the relative motion of contact surfaces that is caused by deformation of a contact body by an external alternating fatigue stress. Fretting increases the tensile and shear stresses in the contact area and causes defects in the contact area, leading to premature crack initiation, as compared to conventional fatigue where no contact relationship exists, and therefore fretting reduces the fatigue strength of the part, leading to a significant reduction in the fatigue life of the part, and even failure of the part below the fatigue limit of the material. For parts subjected to fretting fatigue, the life is reduced by 30% to 80%.
When the influence of a certain factor such as normal load on fretting fatigue performance is researched, a method of changing the factor such as normal load and keeping other factors unchanged is usually adopted, if the fretting fatigue life of a test piece under an unknown normal load needs to be determined, because the existing research does not include an experiment or numerical calculation result under the normal load, the experiment or numerical calculation needs to be carried out again under the unknown normal load, tens of thousands or even tens of thousands of fretting cycles are usually required, 50 factors influencing the fretting process are up to, and once the factors are researched, the time cost and the money cost are greatly increased. If machine learning is carried out on the existing fretting fatigue experiment or numerical calculation data, and then the unknown fretting fatigue performance is predicted, the efficiency of solving the fretting fatigue problem can be greatly improved, and the artificial neural network can be competent for the work.
The artificial neural network resembles the human brain and includes an input layer, a hidden layer, and an output layer. Before the input layer data reaches the output layer, the input layer data needs to be processed by a hidden layer. Except for the output layer, each neuron of the input layer and the hidden layer is multiplied by weight, then added with bias and finally treated by an activation function to be used as a neuron of the next layer.
The following problems exist in the artificial neural network currently used for studying fretting fatigue: in order to obtain good prediction performance, the artificial neural network needs a large amount of experimental data for training and verification, and the cost is too high; in addition, the current artificial neural network can only achieve local optimization and cannot achieve global optimization, and the prediction precision of the artificial neural network is directly influenced.
The invention content is as follows:
the technical problem to be solved by the invention is as follows: the method overcomes the defects of the prior art, obtains the artificial neural network with low cost and global optimization by improving the prior artificial neural network, solves the problems of high cost and incapability of achieving global optimization when the prior artificial neural network is adopted to predict the fretting fatigue performance, and obtains the fretting fatigue performance prediction method based on the artificial neural network with high prediction precision.
The technical scheme of the invention is as follows: a jogging fatigue performance prediction method based on an artificial neural network obtains corresponding data parameters through a series of experiments, constructs a jogging fatigue numerical model and the artificial neural network, optimizes the artificial neural network by using a back propagation algorithm according to errors of a prediction result of the artificial neural network and a calculation result of the jogging fatigue numerical, so that the artificial neural network achieves global optimization, and finally completes accurate prediction of the jogging fatigue performance, and the method comprises the following specific steps:
the method comprises the following steps: carrying out uniaxial tension experiment and axial constant-amplitude fatigue experiment by using a circular-section test piece to obtain the elastic modulus, Poisson's ratio and stress strain curve of the material, and the fatigue strength coefficient and fatigue strength index of the material;
step two: performing fracture mechanics experiments by adopting a compact tension-shear test piece to obtain a material constant in a fatigue crack propagation stage;
step three: performing a fretting fatigue test on an incomplete contact pair adopting a cylindrical pressure head and a flat plate test piece based on a double-vibrator fretting fatigue test device to obtain a crack initiation position, a crack initiation angle, a crack initiation life, a crack propagation path and a crack propagation life of the fretting fatigue test piece;
step four: according to the data obtained in the step one, a finite element method is utilized, and the maximum normal stress range criterion delta sigma based on the critical surface is combinedeq＝Δσn,maxAnd modified Basquin's formula
Establishing a numerical model of a fretting fatigue crack initiation stage, and calculating a fretting fatigue crack initiation position, a crack initiation angle and a crack initiation life based on the model;
wherein, Delta sigmaeqFor equivalent stress range, Δ σn,maxThe maximum normal stress range criterion delta sigma of the critical surface is used as the maximum normal stress rangeeq＝Δσn,maxCalculating the fretting fatigue crack initiation position and the crack initiation angle sigma'fIs the fatigue strength coefficient, σmModified Basquin formula for mean stress, b for fatigue strength index
Calculating fretting fatigue crack initiation life Ni；
Step five: according to the data obtained in the step two, an extension finite element method is utilized, and the maximum normal stress range criterion delta sigma based on the critical surface is combinedeq＝Δσn,maxAnd Paris formula
Establishing a numerical model of the fretting fatigue crack propagation stage, and calculating fretting fatigue crack propagation path and crack propagation based on the modelThe service life is extended;
wherein, Delta sigmaeqFor equivalent stress range, Δ σn,maxThe maximum normal stress range criterion delta sigma of the critical surface is used as the maximum normal stress rangeeq＝Δσn,maxCalculating the micro-motion fatigue crack propagation path, d is a differential sign, a is the crack length, deltaG is the relative fracture energy release rate, c1And c2Paris's formula for fatigue crack propagation stage material constants
Calculating fretting fatigue crack propagation life Np；
Step six: establishing artificial neural network, setting weight
Biasing
And an activation function fiPredicting errors of output layer neurons obtained after hidden layer processing of training set input layer neurons and the micro-motion fatigue numerical calculation results in the fourth step and the fifth step, continuously correcting weights and biases by using a back propagation algorithm based on the errors, and primarily optimizing the artificial neural network, so that the primarily optimized artificial neural network can achieve global optimization for all training set input layer neurons;
step seven: based on the weight obtained in step six
Biasing
And an activation function fiPredicting errors of output layer neurons obtained by processing input layer neurons in the verification set through a hidden layer and the calculation results of the fretting fatigue values in the fourth step and the fifth step, wherein if the errors reach the minimum value, the initially optimized artificial neural network can achieve global optimization for all input layer neurons; otherwise, use the inverseContinuously correcting the weight and the bias in the direction of the propagation algorithm, and finally optimizing the artificial neural network, so that the finally optimized artificial neural network can achieve global optimization for all input layer neurons;
step eight: and predicting the fretting fatigue performance according to the obtained artificial neural network and the existing operation method.
Further, in the fourth step, the accuracy of the numerical model at the fretting fatigue crack initiation stage is verified by using the data obtained in the third step.
Further, in the fifth step, the accuracy of the numerical model at the fretting fatigue crack propagation stage is verified by using the data obtained in the third step.
Further, in the sixth step, the artificial neural network includes an input layer, a hidden layer, and an output layer, and the input layer neurons are processed by a classification algorithm into training set input layer neurons and verification set input layer neurons.
Further, the training set input layer neurons include only the minimum, maximum, and median values of normal, tangential, and distal fatigue loads.
Further, the input layer neurons processed by the classification algorithm enter the output layer after being processed by the hidden layer, and the output layer neurons comprise a micro fatigue crack initiation position, a crack initiation angle, a crack initiation life, a crack propagation path and a crack propagation life.
The invention has the beneficial effects that:
1. according to the invention, the artificial neural network with low cost and global optimization is obtained by improving the existing artificial neural network, and finally the fretting fatigue performance is predicted according to the improved artificial neural network, so that the problems of high cost and incapability of achieving global optimization when the fretting fatigue performance is predicted by adopting the existing artificial neural network are solved, the experiment and numerical calculation cost is reduced, and the prediction precision is improved.
2. The neuron in the input layer of the training set only comprises the minimum value, the maximum value and the intermediate value of the normal load, the tangential load and the far-end fatigue load, so that the experiment and numerical value calculation cost is reduced, and when the finally optimized artificial neural network is subsequently used for predicting the fretting fatigue performance, the numerical value of the neuron in the input layer is positioned between the minimum value and the maximum value, extrapolation is not performed, and the high prediction precision of the artificial neural network is ensured.
3. The invention divides the input layer neurons into training input layer neurons and verification input layer neurons by using a classification algorithm, and according to the error of the obtained output layer neurons and the micro-motion fatigue numerical calculation result, the weights and the offsets are continuously corrected by using a back propagation algorithm, and the artificial neural network is respectively subjected to preliminary optimization and final optimization, so that the preliminary optimization error and the final optimization error can reach the minimum value, the finally optimized artificial neural network can reach the global optimum for all the input layer neurons, and the micro-motion fatigue experiment and numerical calculation cost can be reduced on the premise of improving the prediction precision.
Description of the drawings:
FIG. 1 is a diagram of an improved low-cost and globally optimal artificial neural network.
The specific implementation mode is as follows:
example (b): see fig. 1.
A micromotion fatigue performance prediction method based on an artificial neural network comprises the steps of obtaining a material elastic modulus, a Poisson ratio and a stress strain curve through a uniaxial tensile experiment, obtaining a material fatigue strength coefficient and a fatigue strength index through an axial constant-amplitude fatigue experiment, obtaining a material constant in a fatigue crack propagation stage through a fracture mechanics experiment, and obtaining a crack initiation position, a crack initiation angle, a crack initiation life, a crack propagation path and a crack propagation life of a micromotion fatigue test piece based on the micromotion fatigue experiment;
constructing a fretting fatigue numerical model, establishing fretting fatigue crack initiation stage and propagation stage numerical models by utilizing uniaxial tension experiments, axial constant amplitude fatigue experiments and fracture mechanics experimental data and combining a finite element method and a propagation finite element method, and verifying the accuracy of the fretting fatigue crack initiation stage and propagation stage numerical models by using fretting fatigue experimental data;
establishing a low-cost and global optimal artificial neural network for predicting the fretting fatigue performance, dividing input layer neurons processed by a classification algorithm into training set input layer neurons and verification set input layer neurons, continuously correcting weights and biases by using a back propagation algorithm according to errors of obtained output layer neurons and fretting fatigue numerical calculation results, and respectively performing preliminary optimization and final optimization on the artificial neural network, so that the preliminary optimization errors and the final optimization errors can reach the minimum values, and the finally optimized artificial neural network can reach the global optimal for all the input layer neurons.
The present application will be described in detail below with reference to the drawings and examples.
Wherein the reference numerals in the attached figure 1 have the following meanings:
xi(i ═ 1, 2, …, k) is the ith neuron in the input layer before being processed by the classification algorithm.
yi(i ═ 1, 2, …, p) is the input layer ith neuron processed by the classification algorithm: for the training phase, yi(i-1, 2, …, p) is the ith neuron of the training set input layer, and p-p13; for the verification phase, yi(i ═ 1, 2, …, p) is the ith neuron in the input layer of the verification set, and p ═ p2(ii) a And p is1+p2＝k。
zi(i ═ 1, 2, …, q) is the i-th neuron of the hidden layer.
oi(i ═ 1, 2, …, r) is the output layer ith neuron.
fi(i ═ 1, 2) is the i-th layer activation function.
k is the number of input layer neurons before being processed by the classification algorithm.
And p is the number of input layer neurons processed by the classification algorithm.
p1The number of input layer neurons is the training set.
p2To verify the number of input layer neurons.
s is the layer number.
q is the number of hidden layer neurons.
r is the number of output layer neurons.
The specific steps for predicting the fretting fatigue performance are as follows:
(1) uniaxial tension experiment, axial constant amplitude fatigue experiment, fracture mechanics experiment and fretting fatigue experiment.
(1.1) adopting a circular-section test piece to carry out uniaxial tension test and axial constant-amplitude fatigue test to obtain the elastic modulus E, Poisson ratio v, a stress strain curve and a material fatigue strength coefficient sigma'fAnd a fatigue strength index b.
(1.2) adopting a compact tension-shear test piece to carry out fracture mechanics experiment to obtain a material constant c in a fatigue crack propagation stage1、c2。
(1.3) carrying out a fretting fatigue test on the incomplete contact pair adopting the cylindrical pressure head and the flat plate test piece based on the double-vibrator fretting fatigue test device to obtain the crack initiation position, the crack initiation angle, the crack initiation life, the crack propagation path and the crack propagation life of the fretting fatigue test piece.
(2) And calculating a fretting fatigue value.
(2.1) combining the criterion of maximum normal stress range Δ σ based on the critical plane using the finite element method according to the experimental data of (1.1)eq＝Δσn,maxAnd modified Basquin's formula
Establishing a numerical model of the fretting fatigue crack initiation stage, and calculating the fretting based on the modelFatigue crack initiation position, crack initiation angle, crack initiation life; wherein, Delta sigmaeqFor equivalent stress range, Δ σn,maxThe maximum normal stress range criterion delta sigma of the critical surface is used as the maximum normal stress rangeeq＝Δσn,maxCalculating the fretting fatigue crack initiation position and the crack initiation angle sigma'fIs the fatigue strength coefficient, σmModified Basquin formula for mean stress, b for fatigue strength index
Calculating fretting fatigue crack initiation life Ni(ii) a And (3) verifying the accuracy of the numerical model of the fretting fatigue crack initiation stage by using the experimental data of (1.3).
(2.2) combining the criterion of maximum normal stress range Δ σ based on the critical plane using the extended finite element method according to the experimental data of (1.2)eq＝Δσn,maxAnd Paris formula
Establishing a numerical model of the fretting fatigue crack propagation stage, and calculating a fretting fatigue crack propagation path and a crack propagation life based on the model; wherein, Delta sigmaeqFor equivalent stress range, Δ σn,maxThe maximum normal stress range criterion delta sigma of the critical surface is used as the maximum normal stress rangeeq＝Δσn,maxCalculating the micro-motion fatigue crack propagation path, d is a differential sign, a is the crack length, deltaG is the relative fracture energy release rate, c1And c2Paris's formula for fatigue crack propagation stage material constants
Calculating fretting fatigue crack propagation life Np(ii) a And (3) verifying the accuracy of the numerical model in the fretting fatigue crack propagation stage by using the experimental data of (1.3).
(3) And establishing a low-cost and global optimal artificial neural network for predicting the fretting fatigue performance.
(3.1) the artificial neural network comprises an input layer, a hidden layer and an output layer; the input layer neurons before being processed by the classification algorithm comprise all input parameters, namely all normal loads, all tangential loads and all far-end fatigue loads; the input layer neurons processed by the classification algorithm are divided into training set input layer neurons and verification set input layer neurons; the neuron of the training set input layer only comprises the minimum value, the maximum value and the middle value of the normal load, the tangential load and the far-end fatigue load; the neuron of the input layer processed by the classification algorithm enters the output layer after being processed by the hidden layer; the neuron of the output layer comprises a fretting fatigue crack initiation position, a crack initiation angle, a crack initiation life, a crack propagation path and a crack propagation life.
(3.2) training stage: setting weights
Biasing
And an activation function fiPredicting the error between the output layer neuron obtained after the training set input layer neuron is processed by the hidden layer and the calculation result of the fretting fatigue value in the step (2); based on the error, the weight and the bias are continuously corrected by utilizing a back propagation algorithm, namely, the artificial neural network is preliminarily optimized, so that the preliminarily optimized artificial neural network can achieve global optimization for all the training set input layer neurons.
(3.3) a verification stage: based on weight after training phase
Biasing
And an activation function fiPredicting and verifying the error between the output layer neuron obtained after the input layer neuron is processed by the hidden layer and the calculation result of the fretting fatigue value in the step (2); if the error reaches the minimum value, the initially optimized artificial neural network can achieve global optimization for all input layer neurons; otherwise, continuously correcting the weight by using a back propagation algorithmAnd biasing to enable the errors in the verification stage and the training stage to reach the minimum value, namely, the artificial neural network is finally optimized, so that the finally optimized artificial neural network can achieve global optimization for all input layer neurons.
(4) And predicting the fretting fatigue performance according to the existing operation method according to the improved artificial neural network.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, and all simple modifications, equivalent variations and modifications made to the above embodiment according to the technical spirit of the present invention still fall within the scope of the technical solution of the present invention.
Claims (6)
1. A jogging fatigue performance prediction method based on an artificial neural network obtains corresponding data parameters through a series of experiments, constructs a jogging fatigue numerical model and the artificial neural network, optimizes the artificial neural network by using a back propagation algorithm according to errors of a prediction result of the artificial neural network and a calculation result of the jogging fatigue numerical, so that the artificial neural network achieves global optimization, and finally completes accurate prediction of the jogging fatigue performance, and the method comprises the following specific steps:
the method comprises the following steps: carrying out uniaxial tension experiment and axial constant-amplitude fatigue experiment by using a circular-section test piece to obtain the elastic modulus, Poisson's ratio and stress strain curve of the material, and the fatigue strength coefficient and fatigue strength index of the material;
step two: performing fracture mechanics experiments by adopting a compact tension-shear test piece to obtain a material constant in a fatigue crack propagation stage;
step three: performing a fretting fatigue test on an incomplete contact pair adopting a cylindrical pressure head and a flat plate test piece based on a double-vibrator fretting fatigue test device to obtain a crack initiation position, a crack initiation angle, a crack initiation life, a crack propagation path and a crack propagation life of the fretting fatigue test piece;
step four: according to the data obtained in the step one, a finite element method is utilized, and the maximum normal stress range criterion delta sigma based on the critical surface is combinedeq＝Δσn,maxAnd modified Basquin formula
Establishing a numerical model of a fretting fatigue crack initiation stage, and calculating a fretting fatigue crack initiation position, a crack initiation angle and a crack initiation life based on the model;
wherein, Delta sigmaeqFor equivalent stress range, Δ σn,maxThe maximum normal stress range criterion delta sigma of the critical surface is used as the maximum normal stress rangeeq＝Δσn,maxCalculating the fretting fatigue crack initiation position and the crack initiation angle sigma'fIs the fatigue strength coefficient, σmModified Basquin formula for mean stress, b for fatigue strength index
Calculating fretting fatigue crack initiation life Ni；
Step five: according to the data obtained in the step two, an extension finite element method is utilized, and the maximum normal stress range criterion delta sigma based on the critical surface is combinedeq＝Δσn,maxAnd Paris formula
Establishing a numerical model of the fretting fatigue crack propagation stage, and calculating a fretting fatigue crack propagation path and a crack propagation life based on the model;
wherein, Delta sigmaeqFor equivalent stress range, Δ σn,maxThe maximum normal stress range criterion delta sigma of the critical surface is used as the maximum normal stress rangeeq＝Δσn,maxCalculating the micro-motion fatigue crack propagation path, d is a differential sign, a is the crack length, deltaG is the relative fracture energy release rate, c1And c2Paris's formula for fatigue crack propagation stage material constants
Calculating fretting fatigue crack propagation life Np；
Step six: building (2)Setting weights by setting artificial neural network
Biasing
And an activation function fiPredicting errors of output layer neurons obtained after hidden layer processing of training set input layer neurons and the micro-motion fatigue numerical calculation results in the fourth step and the fifth step, continuously correcting weights and biases by using a back propagation algorithm based on the errors, and primarily optimizing the artificial neural network, so that the primarily optimized artificial neural network can achieve global optimization for all training set input layer neurons;
step seven: based on the weight obtained in step six
Biasing
And an activation function fiPredicting errors of output layer neurons obtained by processing input layer neurons in the verification set through a hidden layer and the calculation results of the fretting fatigue values in the fourth step and the fifth step, wherein if the errors reach the minimum value, the initially optimized artificial neural network can achieve global optimization for all input layer neurons; otherwise, continuously correcting the weight and the bias by using a back propagation algorithm, and finally optimizing the artificial neural network, so that the finally optimized artificial neural network can achieve global optimization for all input layer neurons;
step eight: and predicting the fretting fatigue performance according to the obtained artificial neural network and the existing operation method.
2. The method for predicting fretting fatigue performance based on artificial neural network as claimed in claim 1, wherein: and in the fourth step, the accuracy of the numerical model at the initiation stage of the fretting fatigue crack is verified by using the data obtained in the third step.
3. The method for predicting fretting fatigue performance based on artificial neural network as claimed in claim 1, wherein: and in the fifth step, the data obtained in the third step are used for verifying the accuracy of the numerical model in the fretting fatigue crack propagation stage.
4. The method for predicting fretting fatigue performance based on artificial neural network as claimed in claim 1, wherein: in the sixth step, the artificial neural network comprises an input layer, a hidden layer and an output layer, and the input layer neurons are processed into training set input layer neurons and verification set input layer neurons through a classification algorithm.
5. The method for predicting fretting fatigue performance based on artificial neural network as claimed in claim 4, wherein: the training set input layer neurons include only the minimum, maximum, and median values of normal, tangential, and distal fatigue loads.
6. The method for predicting fretting fatigue performance based on artificial neural network as claimed in claim 1, wherein: the neuron of the input layer processed by the classification algorithm enters the output layer after being processed by the hidden layer, and the neuron of the output layer comprises a micro fatigue crack initiation position, a crack initiation angle, a crack initiation life, a crack propagation path and a crack propagation life.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010929127.7A CN112052615B (en) | 2020-09-07 | 2020-09-07 | Micro fatigue performance prediction method based on artificial neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010929127.7A CN112052615B (en) | 2020-09-07 | 2020-09-07 | Micro fatigue performance prediction method based on artificial neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112052615A true CN112052615A (en) | 2020-12-08 |
| CN112052615B CN112052615B (en) | 2023-05-09 |
Family
ID=73609900
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010929127.7A Active CN112052615B (en) | 2020-09-07 | 2020-09-07 | Micro fatigue performance prediction method based on artificial neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112052615B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112784495A (en) * | 2021-01-28 | 2021-05-11 | 郑州轻工业大学 | Mechanical structure real-time fatigue life prediction method based on data driving |
| CN115169694A (en) * | 2022-07-06 | 2022-10-11 | 天津大学 | Method for realizing dynamic propagation and service life prediction of subcritical crack |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090276166A1 (en) * | 2008-05-05 | 2009-11-05 | Qigui Wang | Methods and systems to predict fatigue life in aluminum castings |
| CN102693450A (en) * | 2012-05-16 | 2012-09-26 | 北京理工大学 | A prediction method for crankshaft fatigue life based on genetic nerve network |
| CN103105406A (en) * | 2011-11-09 | 2013-05-15 | 北京有色金属研究总院 | Method for observing crack propagation path of titanium alloy under plane strain state |
| CN108984926A (en) * | 2018-07-25 | 2018-12-11 | 湖南大学 | A kind of Multiaxial Fatigue Life Prediction method |
| CN110967267A (en) * | 2019-11-25 | 2020-04-07 | 中国民用航空飞行学院 | Test method for judging fatigue crack initiation life |
-
2020
- 2020-09-07 CN CN202010929127.7A patent/CN112052615B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090276166A1 (en) * | 2008-05-05 | 2009-11-05 | Qigui Wang | Methods and systems to predict fatigue life in aluminum castings |
| CN103105406A (en) * | 2011-11-09 | 2013-05-15 | 北京有色金属研究总院 | Method for observing crack propagation path of titanium alloy under plane strain state |
| CN102693450A (en) * | 2012-05-16 | 2012-09-26 | 北京理工大学 | A prediction method for crankshaft fatigue life based on genetic nerve network |
| CN108984926A (en) * | 2018-07-25 | 2018-12-11 | 湖南大学 | A kind of Multiaxial Fatigue Life Prediction method |
| CN110967267A (en) * | 2019-11-25 | 2020-04-07 | 中国民用航空飞行学院 | Test method for judging fatigue crack initiation life |
Non-Patent Citations (3)
| Title |
|---|
| 周宇 等: ""考虑磨耗的钢轨疲劳裂纹萌生寿命预测仿真"", 《铁道学报》 * |
| 张华阳: ""残余应力对发动机机体隔板微动疲劳性能影响规律的研究"", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 蒋钦贤: ""超弹性NiTi合金断裂行为实验和有限元模拟"", 《中国优秀硕士学位论文全文数据库 基础科学辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112784495A (en) * | 2021-01-28 | 2021-05-11 | 郑州轻工业大学 | Mechanical structure real-time fatigue life prediction method based on data driving |
| CN112784495B (en) * | 2021-01-28 | 2021-09-24 | 郑州轻工业大学 | Mechanical structure real-time fatigue life prediction method based on data driving |
| CN115169694A (en) * | 2022-07-06 | 2022-10-11 | 天津大学 | Method for realizing dynamic propagation and service life prediction of subcritical crack |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112052615B (en) | 2023-05-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yang et al. | Machine learning and artificial neural network prediction of interfacial thermal resistance between graphene and hexagonal boron nitride | |
| Abouhamze et al. | Multi-objective stacking sequence optimization of laminated cylindrical panels using a genetic algorithm and neural networks | |
| Zhou et al. | Machine learning‐based genetic feature identification and fatigue life prediction | |
| CN112052615A (en) | Micro-motion fatigue performance prediction method based on artificial neural network | |
| CN109255202A (en) | A kind of predictor method for mechanical component fatigue crack initiation life | |
| Bezerra et al. | Artificial neural networks applied to epoxy composites reinforced with carbon and E-glass fibers: Analysis of the shear mechanical properties | |
| Kumar et al. | Fatigue life prediction of glass fiber reinforced epoxy composites using artificial neural networks | |
| CN112784495B (en) | Mechanical structure real-time fatigue life prediction method based on data driving | |
| CN105022898A (en) | Crack performance measurement and optimization solution method for composite material glue joint structure | |
| Altabey et al. | Fatigue life prediction for carbon fibre/epoxy laminate composites under spectrum loading using two different neural network architectures | |
| CN109308004B (en) | Real triaxial rigidity testing machine post-peak loading process servo control system and method | |
| Zhang et al. | Embedding artificial neural networks into twin cohesive zone models for composites fatigue delamination prediction under various stress ratios and mode mixities | |
| Ly et al. | Multi-objective optimization of laminated functionally graded carbon nanotube-reinforced composite plates using deep feedforward neural networks-NSGAII algorithm | |
| Ding et al. | Jaya-based long short-term memory neural network for structural damage identification with consideration of measurement uncertainties | |
| CN114201911A (en) | Rubber material fatigue life prediction method based on extreme learning machine | |
| Lyu et al. | Machine learning meta-models for fast parameter identification of the lattice discrete particle model | |
| Tran et al. | Optimization design of laminated functionally carbon nanotube-reinforced composite plates using deep neural networks and differential evolution | |
| Deng et al. | On-line surface roughness classification for multiple CNC milling conditions based on transfer learning and neural network | |
| Mohanty et al. | Prediction of residual fatigue life under interspersed mixed‐mode (I and II) overloads by Artificial Neural Network | |
| Elenchezhian et al. | Damage precursor identification in composite laminates using data driven approach | |
| Most | Approximation of complex nonlinear functions by means of neural networks | |
| Jam et al. | Free vibrations of three-parameter functionally graded plates resting on pasternak foundations | |
| Wang et al. | Improved back propagation neural network based on the enrichment for the crack propagation | |
| Wahab | A data-assisted physics informed neural network (Da-Pinn) for fretting fatigue lifetime prediction | |
| Yu et al. | An artificial neural network model for flexoelectric actuation and control of beams |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |