Method for loading axial stress and circumferential stress of metal pipe based on neural network

Abstract

The invention relates to a method for loading axial stress and circumferential stress of a metal pipe based on a neural network, and belongs to the technical field of multi-axial stress measurement of metal pipe fittings. The method comprises the steps that firstly, a metal pipe bulging model is established, and the stress ratio of the axial stress to the circumferential stress generated by multiple axial displacement and hydraulic expansive force to the pipe is obtained; further establishing a stress ratio data set; secondly, obtaining a metal pipe loading path prediction model based on the stress ratio data set; and finally, inputting the tested stress ratio and hydraulic expansive force into a metal pipe loading path prediction model, and outputting predicted axial displacement. The predicted axial displacement is input into the pipe liquid expansion testing machine, the axial stress and the circumferential stress of the metal pipe are obtained based on the set stress ratio and the hydraulic expansion force, and the problem that the testing cost of the pipe liquid expansion testing machine provided with a complex feedback control system is high is solved.

Description

Technical Field

[0001] The invention relates to the technical field of multi-axial stress measurement of metal pipes, and in particular to a method for loading axial stress and circumferential stress of metal pipes based on a neural network. Background Art

[0002] Sheet metal plastic forming technology has long been widely used, especially for alloys used in aircraft and spacecraft. Traditional sheet metal testing has high test time and mold processing costs, making finite element analysis increasingly important in calculating the deformation behavior of sheet metal and predicting unstable deformation models such as fracture or wrinkling. The accurate prediction of these models is based on constitutive models obtained from accurate test data. In the sheet metal forming process, material properties can be measured by yield strength, tensile strength and work hardening coefficient, which are usually obtained from standard uniaxial tensile tests.

[0003] However, sheet metal parts of automobiles, ships and aircraft are usually in a complex stress state, which makes the metal sheet subject to multi-axial loading conditions. If the performance of the material obtained by uniaxial tensile test is equivalent to the performance of biaxial tensile test, there will be a large error. Therefore, the multi-axial stress and strain in the plastic deformation stage cannot be measured by traditional uniaxial test but need to be measured by biaxial loading test. As the research deepens, scholars' research on the subsequent yield and strengthening behavior of test materials also requires theoretical verification of the performance data of materials under large strain.

[0004] The study of material properties under large strain multi-axial conditions is usually conducted in the form of pipes. This test requires a pipe liquid expansion testing machine equipped with a complex feedback control system. By controlling the pipe liquid expansion force (internal pressure) and the pipe lateral displacement (axial) feed rate of each increment during the expansion process, a linear stress path is generated in the center area of ​​the pipe, which makes the cost of each test relatively expensive. Summary of the invention

[0005] In view of the above problems, the present invention provides a method for loading axial stress and circumferential stress of metal pipes based on neural network. The present invention first establishes a metal pipe expansion model, obtains stress ratios of axial stress and circumferential stress generated by multiple axial displacements and hydraulic expansion forces on the pipe, and further establishes a stress ratio data set; secondly, a metal pipe loading path prediction model is obtained based on the stress ratio data set; finally, the stress ratio and hydraulic expansion force described in the test are input into the metal pipe loading path prediction model, and the predicted axial displacement is output; the predicted axial displacement is input into a pipe liquid expansion testing machine, and the axial stress and circumferential stress of the metal pipe are obtained based on the set stress ratio and hydraulic expansion force, thereby solving the problem of expensive testing costs of pipe liquid expansion testing machines equipped with complex feedback control systems.

[0006] The present invention provides a method for loading axial stress and circumferential stress of a metal pipe based on a neural network, comprising:

[0007] Step 1: Establish a metal tube bulging model using finite element software; obtain multiple hydraulic expansion forces and multiple axial displacements; obtain a stress ratio data set based on the multiple hydraulic expansion forces, multiple axial displacements and the metal tube bulging model;

[0008] Preferably, the specific steps of obtaining the stress ratio data set in step 1 include:

[0009] Step 1.1, using Abaqus finite element software to establish a metal tube bulging model;

[0010] Furthermore, the metal tube bulging model comprises a metal tube 1, a hollow cylindrical fixture 1 2 and a hollow cylindrical fixture 2 3;

[0011] The hollow cylindrical clamp 1 2 and the hollow cylindrical clamp 2 3 are respectively connected to the two ends of the metal pipe according to binding constraints.

[0012] Step 1.2, based on test experience, obtain multiple hydraulic expansion forces and multiple axial displacements;

[0013] Step 1.3, combining each axial displacement and each pipe hydraulic expansion force in pairs as a loading path combination, to obtain multiple groups of loading path combinations;

[0014] Step 1.4, inputting each group of loading path combinations into the metal tube bulging model, respectively, obtaining the axial stress and circumferential stress of the metal tube bulging area and the intermediate unit generated in the metal tube bulging area, and obtaining the stress ratio corresponding to the axial stress and the circumferential stress; based on the axial stress, circumferential stress and the corresponding stress ratio corresponding to each group of loading path combinations, a stress ratio data set is established;

[0015] Step 2, building a neural network model; the input of the neural network model is hydraulic expansion force and stress ratio; the output is axial displacement;

[0016] Exemplarily, the neural network model includes an input layer, two hidden layers and an output layer; an activation function sigmoid layer is set between the two hidden layers, and each layer is set with a number of neurons, and the neurons between layers are interconnected;

[0017] Step 3: training the neural network model based on the stress ratio data set to obtain a metal pipe loading path prediction model;

[0018] Preferably, the specific steps of obtaining the metal pipe loading path prediction model include:

[0019] Step 3.1. Normalize the stress ratio data set to obtain a processed stress ratio data set; initialize the weights and thresholds of the neural network;

[0020] Step 3.2. When k = 1, it indicates the first hydraulic expansion force, stress ratio and axial displacement;

[0021] Select the kth hydraulic expansion force and stress ratio in the processed stress ratio data set;

[0022] Step 3.3. Input the k-th hydraulic expansion force and stress ratio into the neural network model A k , obtain the kth actual axial displacement; obtain the kth axial displacement corresponding to the kth hydraulic expansion force and stress ratio based on the stress ratio data set;

[0023] Step 3.4. Compare the k-th actual axial displacement with the k-th axial displacement to see if there is an error; if so, adjust the neural network model A based on the error. k The weights are updated to obtain the neural network model A k+1 , go to the next step; if it does not exist, go directly to the next step;

[0024] Step 3.5. Determine whether k is greater than or equal to K, where K is the number of hydraulic expansion forces in the stress ratio data set. If so, the training ends and the neural network model A is obtained. k+1 , the neural network model A k+1 as a metal pipe loading path prediction model; if not, let k = k + 1 and return to step 3.3;

[0025] Furthermore, the metal pipe loading path prediction model expression is:

[0026]

[0027] Among them, y k is the predicted kth axial displacement, w jk is the weight of the k-th axial displacement in the j-th hidden layer neuron, j = 1, 2, 3...J, J represents the number of hidden layers, a j is the bias weight of the hidden layer, w ij is the weight between the i-th input layer and the j-th hidden layer, b c is the bias weight on the output layer; sig(·) is the sigmoid activation function; x i,k A is the neural network model k The input value of input layer i, i = 1, 2, 3...I, I represents the number of input layers; y k is the kth actual axial displacement.

[0028] Step 4: Set the stress ratio and hydraulic expansion force according to the test conditions;

[0029] Inputting the set stress ratio and hydraulic expansion force into a metal pipe loading path prediction model, and outputting a predicted axial displacement;

[0030] The predicted axial displacement is input into a pipe hydraulic expansion testing machine, and the axial stress and circumferential stress of the metal pipe are obtained based on the set stress ratio and hydraulic expansion force.

[0031] Preferably, the expressions of the axial stress and circumferential stress of the metal pipe are:

[0032]

[0033] Among them, σ is stress, ε is strain, E is the elastic modulus of the material, k is the strength coefficient of the material, and n is the hardening exponent of the material.

[0034] Compared with the prior art, the present invention has at least the following beneficial effects:

[0035] The present invention only needs to generate the required boundary conditions, axial thrust displacement and pressure distribution by numerical simulation, and no longer needs to use the complex and expensive control system on the pipe multi-axis testing machine, which can effectively reduce the cost of the testing machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings are only for the purpose of illustrating particular embodiments and are not to be construed as limiting the invention.

[0037] Figure 1 It is a schematic diagram of a flow chart of loading axial stress and circumferential stress of a metal pipe based on a neural network in an embodiment of the present invention;

[0038] Figure 2 is a schematic diagram of a front view of a metal pipe sample in an embodiment of the present invention;

[0039] Figure 3 A schematic diagram of a metal tube bulging model in an embodiment of the present invention;

[0040] Figure 4 Schematic diagram of a neural network in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. In addition, the present invention can also be implemented in other ways different from those described herein, and therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0042] A specific embodiment of the present invention, as Figure 1-4 , discloses a method for loading axial stress and circumferential stress of metal pipes based on neural network. In order to illustrate the effectiveness of the method proposed by the present invention, the above technical solution of the present invention is described in detail through a specific embodiment below. The specific implementation steps are as follows:

[0043] Step 1: Establish a metal tube bulging model using finite element software; obtain multiple hydraulic expansion forces and multiple axial displacements; obtain a stress ratio data set based on the multiple hydraulic expansion forces, multiple axial displacements and the metal tube bulging model;

[0044] In one embodiment of the present invention, the specific steps of obtaining the stress ratio data set in step 1 include:

[0045] Step 1.1, using Abaqus finite element software to establish a metal tube bulging model;

[0046] Furthermore, the metal tube bulging model comprises a metal tube 1, a hollow cylindrical fixture 1 2 and a hollow cylindrical fixture 2 3;

[0047] The hollow cylindrical clamp 1 2 and the hollow cylindrical clamp 2 3 are respectively connected to the two ends of the metal pipe according to binding constraints.

[0048] The binding constraints are formulated based on Abaqus / CAE finite element software;

[0049] The axial displacement is the displacement of the two hollow cylindrical clamps moving axially toward the middle position of the metal pipe.

[0050] Exemplarily, the bulging area of ​​the metal tube is located in the middle, and the length of the bulging area is twice the outer diameter of the tube;

[0051] like Figure 3 , the length of the bulging area is 180mm;

[0052] The metal pipe is a 6061 aluminum alloy O-state pipe annealed at 480°C for 5 hours and air-cooled;

[0053] The length of the metal tube used is 260mm, the outer diameter is 60mm, and the wall thickness is 2mm; the clamps installed at the left and right ends are 40mm respectively, so the bulging area is 180mm;

[0054] Step 1.2, based on test experience, obtain multiple hydraulic expansion forces and multiple axial displacements;

[0055] Step 1.3, combining each axial displacement and each pipe hydraulic expansion force in pairs as a loading path combination, to obtain multiple groups of loading path combinations;

[0056] Step 1.4, inputting each group of loading path combinations into the metal tube bulging model, respectively, obtaining the axial stress and circumferential stress of the metal tube bulging area and the intermediate unit generated in the metal tube bulging area, and obtaining the stress ratio corresponding to the axial stress and the circumferential stress; based on the axial stress, circumferential stress and the corresponding stress ratio corresponding to each group of loading path combinations, a stress ratio data set is established;

[0057] Step 2, building a neural network model; the input of the neural network model is hydraulic expansion force and stress ratio; the output is axial displacement;

[0058] Exemplarily, the neural network model includes an input layer, two hidden layers and an output layer; an activation function sigmoid layer is set between the two hidden layers, and each layer is set with a number of neurons, and the neurons between layers are interconnected;

[0059] Exemplarily, there are 15 hidden layer nodes, the training set accounts for 80%, the maximum number of iterations is set to 1000 times, the error threshold is 1e-6, and the learning rate is 0.01.

[0060] Step 3: training the neural network model based on the stress ratio data set to obtain a metal pipe loading path prediction model;

[0061] Preferably, the specific steps of obtaining the metal pipe loading path prediction model include:

[0062] Step 3.1. Normalize the stress ratio data set to obtain a processed stress ratio data set; initialize the weights and thresholds of the neural network;

[0063] Step 3.2. When k = 1, it indicates the first hydraulic expansion force, stress ratio and axial displacement;

[0064] Select the kth hydraulic expansion force and stress ratio in the processed stress ratio data set;

[0065] Step 3.3. Input the k-th hydraulic expansion force and stress ratio into the neural network model A k , obtain the kth actual axial displacement; obtain the kth axial displacement corresponding to the kth hydraulic expansion force and stress ratio based on the stress ratio data set;

[0066] Step 3.4. Compare the k-th actual axial displacement with the k-th axial displacement to see if there is an error; if so, adjust the neural network model A based on the error. k The weights are updated to obtain the neural network model A k+1 , go to the next step; if it does not exist, go directly to the next step;

[0067] Step 3.5. Determine whether k is greater than or equal to K, where K is the number of hydraulic expansion forces in the stress ratio data set. If so, the training ends and the neural network model A is obtained. k+1 , the neural network model A k+1 as a metal pipe loading path prediction model; if not, let k = k + 1 and return to step 3.3;

[0068] Furthermore, the metal pipe loading path prediction model expression is:

[0069]

[0070] Among them, y k is the predicted kth axial displacement, w jk is the weight of the k-th axial displacement in the j-th hidden layer neuron, j = 1, 2, 3...J, J represents the number of hidden layers, a j is the bias weight of the hidden layer, w ij is the weight between the i-th input layer and the j-th hidden layer, b c is the bias weight on the output layer; sig(·) is the sigmoid activation function; x i,k A is the neural network model k The input value of input layer i, i = 1, 2, 3...I, I represents the number of input layers; y k is the kth actual axial displacement.

[0071] Step 4: Set the stress ratio and hydraulic expansion force according to the test conditions;

[0072] Inputting the set stress ratio and hydraulic expansion force into a metal pipe loading path prediction model, and outputting a predicted axial displacement;

[0073] The predicted axial displacement is input into a pipe hydraulic expansion testing machine, and the axial stress and circumferential stress of the metal pipe are obtained based on the set stress ratio and hydraulic expansion force.

[0074] Preferably, the expressions of the axial stress and circumferential stress of the metal pipe are:

[0075]

[0076] Among them, σ is stress, ε is strain, E is the elastic modulus of the material, N is the strength coefficient of the material, and n is the hardening exponent of the material.

[0077] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for loading axial stress and circumferential stress of metal pipes based on neural network, characterized in that: include: Step 1: Establish a metal tube expansion model using finite element software; obtain multiple hydraulic expansion forces and multiple axial displacements; Obtaining a stress ratio data set based on the multiple hydraulic expansion forces, the multiple axial displacements and the metal tube expansion model; Step 2, building a neural network model; the input of the neural network model is hydraulic expansion force and stress ratio; the output is axial displacement; Step 3: training the neural network model based on the stress ratio data set to obtain a metal pipe loading path prediction model; Step 4: setting the stress ratio and hydraulic expansion force according to the test conditions; inputting the set stress ratio and hydraulic expansion force into the metal pipe loading path prediction model, and outputting the predicted axial displacement; The predicted axial displacement is input into a pipe hydraulic expansion testing machine, and the axial stress and circumferential stress of the metal pipe are obtained based on the set stress ratio and hydraulic expansion force.

2. The method for applying axial stress and circumferential stress to a metal pipe according to claim 1, characterized in that: The specific steps of obtaining the stress ratio data set in step 1 include: Step 1.1, using Abaqus finite element software to establish a metal tube bulging model; Step 1.2, based on test experience, obtain multiple hydraulic expansion forces and multiple axial displacements; Step 1.3, combining each axial displacement and each pipe hydraulic expansion force in pairs as a loading path combination, to obtain multiple groups of loading path combinations; Step 1.4, input each group of loading path combinations into the metal tube bulging model, respectively, obtain the axial stress and circumferential stress of the metal tube bulging area and the intermediate unit generated in the metal tube bulging area, and obtain the stress ratio corresponding to the axial stress and the circumferential stress; based on the axial stress, circumferential stress and the corresponding stress ratio corresponding to each group of loading path combinations, establish a stress ratio data set.

3. The method for applying axial stress and circumferential stress to a metal pipe according to claim 1, characterized in that: The neural network model in step 2 includes an input layer, two hidden layers and an output layer; an activation function sigmoid layer is set between the two hidden layers, and each layer is set with a number of neurons, and the neurons between layers are interconnected.

4. The method for applying axial stress and circumferential stress to a metal pipe according to claim 1, characterized in that: The specific steps of obtaining the metal pipe loading path prediction model include: Step 3.

1. Normalize the stress ratio data set to obtain a processed stress ratio data set; initialize the weights and thresholds of the neural network; Step 3.

2. Let k = 1. When k = 1, it represents the first hydraulic expansion force, stress ratio and axial displacement; Select the kth hydraulic expansion force and stress ratio in the processed stress ratio data set; Step 3.

3. Input the k-th hydraulic expansion force and stress ratio into the neural network model A k , obtain the kth actual axial displacement; obtain the kth axial displacement corresponding to the kth hydraulic expansion force and stress ratio based on the stress ratio data set; Step 3.

4. Compare the k-th actual axial displacement with the k-th axial displacement to see if there is an error; if so, adjust the neural network model A based on the error. k The weights are updated to obtain the neural network model A k+1 , go to the next step; if it does not exist, go directly to the next step; Step 3.

5. Determine whether k is greater than or equal to K, where K is the number of hydraulic expansion forces in the stress ratio data set. If so, the training ends and the neural network model A is obtained. k+1 , the neural network model A k+1 As the metal pipe loading path prediction model; if not, let k = k + 1 and return to step 3.

3.

5. The method for applying axial stress and circumferential stress to a metal pipe according to claim 4, characterized in that: The metal pipe loading path prediction model expression is: Among them, y k is the predicted kth axial displacement, w jk is the weight of the k-th axial displacement in the j-th hidden layer neuron, j = 1, 2, 3...J, J represents the number of hidden layers, a j is the bias weight of the hidden layer, w ij is the weight between the i-th input layer and the j-th hidden layer, b c is the bias weight on the output layer; sig(·) is the sigmoid activation function; x i,k A is the neural network model k The input value of input layer i, i = 1, 2, 3...I, I represents the number of input layers; y k is the kth actual axial displacement.

6. The method for applying axial stress and circumferential stress to a metal pipe according to claim 1, characterized in that: The expressions of the axial stress and circumferential stress of the metal pipe are: Among them, σ is stress, ε is strain, E is the elastic modulus of the material, N is the strength coefficient of the material, and n is the hardening exponent of the material.

7. The method for applying axial stress and circumferential stress to a metal pipe according to claim 2, characterized in that: The metal tube bulging model comprises a metal tube (1), a hollow cylindrical fixture 1 (2) and a hollow cylindrical fixture 2 (3); The hollow cylindrical clamp 1 (2) and the hollow cylindrical clamp 2 (3) are respectively connected to the two ends of the metal pipe (1) according to binding constraints.

8. The method for applying axial stress and circumferential stress to a metal pipe according to claim 7, characterized in that: The ratio between the length of the metal tube bulging area and the outer diameter of the metal tube (1) is 1:

2.

9. The method for applying axial stress and circumferential stress to a metal pipe according to claim 7, characterized in that: The binding constraints are formulated based on Abaqus / CAE finite element software.

10. The method for applying axial stress and circumferential stress to a metal pipe according to claim 1, characterized in that: The metal pipe is a 6061 aluminum alloy O-state pipe.

Images