Titanium alloy surface stress prediction method based on subtraction average optimization algorithm

Abstract

According to the titanium alloy surface stress prediction method based on the subtraction average optimization algorithm, the characteristics of high efficiency, high adaptive capacity, small noise influence, high adaptive capacity and the like of the subtraction average optimization algorithm are utilized, so that the accuracy and the applicability of the prediction model are improved. Meanwhile, experimental data and simulation data are transmitted to the BP neural network for training, the number of tests can be reduced, the time and cost of the prediction process are greatly reduced, more experimental sample data are used for training the neural network, and compared with a traditional method that only experimental data are used for training the neural network, the method has the advantages that the training efficiency is improved. The precision and efficiency of predicting the residual stress on the surface of the titanium alloy can be effectively improved, and data support is provided for material design and improvement of a machining process.

Description

Technical Field

[0001] The present invention relates to the field of metrology and detection technology, and more specifically, to a titanium alloy surface stress prediction method based on a subtraction average optimization algorithm. Background Art

[0002] In the field of high-end manufacturing, titanium alloys are widely used in the manufacture of key structural components due to their excellent strength, corrosion resistance and low density. However, high temperature treatment and machining during the manufacturing process often introduce residual stress inside the material, which may have a negative impact on the performance and reliability of the components.

[0003] In the existing technology, the prediction of residual stress of titanium alloy usually relies on direct measurement or empirical estimation. These methods are not only time-consuming and costly, but also often invasive and may damage the test sample. In addition, the traditional prediction model is not accurate and stable enough when dealing with complex variable relationships, such as the interaction between stress and microstructure, and it is difficult to adapt to the real-time monitoring and control needs in the manufacturing process of high-performance titanium alloy parts. Neural network algorithms have been used to improve the accuracy and precision of predictions.

[0004] Patent document CN117951933A discloses a method for predicting the surface stress of titanium alloy based on the dung beetle optimization algorithm. The BP neural network optimized by the dung beetle optimization algorithm can learn deeper nonlinear relationships from finite element simulation data, thereby improving the accuracy and applicability of the prediction model. The dung beetle optimization algorithm (DBOA) is a swarm intelligence optimization algorithm based on the behavior of dung beetles. Although it performs well in some applications, it also has some disadvantages: the dung beetle optimization algorithm may fall into a local optimal solution, especially in a complex search space, resulting in the inability to find a global optimal solution. In some cases, the convergence speed of the algorithm may be slow, especially in high-dimensional problems, and more iterations may be required to achieve satisfactory results. The performance of the algorithm may be very sensitive to parameter settings, and different parameter selections may lead to completely different optimization results. When dealing with large-scale problems, the computational complexity of the dung beetle optimization algorithm may be high, resulting in increased calculation time. Compared with some classic optimization algorithms, the theoretical basis and convergence analysis of the dung beetle optimization algorithm are relatively less, which may affect its application in some fields. In a dynamic environment, the dung beetle optimization algorithm may not be able to quickly adapt to environmental changes, resulting in a decrease in optimization effect. These shortcomings may limit the effectiveness and efficiency of the dung beetle optimization algorithm in some practical applications.

[0005] The current difficulty and focus is how to use neural network algorithms to efficiently and quickly converge neural algorithms in a dynamic environment, improve efficiency and applicability, and improve structural accuracy. Summary of the invention

[0006] In view of this, the present invention aims to provide a titanium alloy surface stress prediction method based on a subtraction average optimization algorithm, comprising the following steps:

[0007] Step 1, establishing a finite element layer-by-layer electrolysis analysis model, and obtaining surface stress simulation data through electrolysis simulation analysis;

[0008] Step 2, electrolyzing the titanium alloy layer by layer to obtain surface stress experimental data of each layer after electrolysis;

[0009] Step 3, verifying the electrolytic analysis model through experimental data and further optimizing the electrolytic analysis model;

[0010] Step 4, identifying and correcting the surface stress experimental data to obtain optimized surface stress experimental data;

[0011] Step 5, using the subtraction average optimization algorithm to optimize the weights and biases of the BP neural network;

[0012] Step 6, based on the optimized BP neural network, training is performed using the optimized surface stress experimental data and surface stress simulation data to predict the surface stress of the titanium alloy.

[0013] Furthermore, in step 1, when simulating the generation and distribution of stress of titanium alloy during processing based on the finite element method, boundary conditions and initial conditions are set.

[0014] Furthermore, the boundary conditions include fixed supports, loading paths and fine grid divisions; and the initial conditions include current density and temperature.

[0015] Furthermore, in step 2, the surface stress experimental data of each layer after electrolysis is obtained, specifically, the surface stress data of each layer after electrolysis is measured by using X-ray diffraction method.

[0016] Furthermore, in step 4, the data set is identified and corrected, including: using the DBSCAN algorithm to identify abnormal data.

[0017] Furthermore, the specific steps of using the DBSCAN algorithm to identify abnormal data are as follows: input a data set, randomly select an unvisited data point p from the data set to start clustering, use p as the access point, and start searching whether the Eps neighborhood of p contains Minpts data points. If so, p is marked as a core point, and a new cluster M is created for the core point p, and all points in the Eps neighborhood are added to the candidate set N; for all objects q that have not been selected in the candidate set N, check whether their Eps neighborhood contains at least Minpts data points. If so, these points are added to N. If q in the candidate set N is not added to any cluster, q is added to the new cluster M, and the unprocessed data points in the candidate set N are continued to be checked. When all points in the candidate set N have been processed, stop, and repeat the above steps until all points have been visited. At this time, any points that have not been added to any cluster are classified as abnormal data.

[0018] Furthermore, in step 5, the steps of optimizing the weights and biases of the BP neural network using the subtraction mean optimization algorithm are as follows:

[0019] S51, randomly initialize the weights and biases of the neural network;

[0020] S52, calculate the loss based on the network output and the true label;

[0021] S53, calculate the gradient of loss with respect to weights and biases;

[0022] S54, update weights and biases;

[0023] S55, repeating S52 to S54 until the loss function converges or reaches a preset number of iterations;

[0024] S56, output the results and return the optimized weights and biases.

[0025] Furthermore, the specific steps of updating the weights and biases in step S54 are to calculate the average values ​​of the current weights and biases, and update the weights and biases using the subtraction mean method. For each weight w i and bias b j The update formula is:

[0026] Among them, η is the learning rate, α is the weight of subtractive averaging, and w avg and b avg is the average of the current weights and biases.

[0027] Furthermore, a finite element layer-by-layer electrolysis analysis model is established, including:

[0028] S11, model construction: 3D model of the alloy was constructed using finite element software, and the alloy was divided into multiple layers, with different electrolysis conditions and material properties considered for each layer;

[0029] S12, electrolysis process simulation: set the chemical composition and current density of the electrolyte, and calculate the electrolysis reaction rate and thermal effect of each layer by layer analysis;

[0030] S13, Stress Analysis: Using linear elastic theory, the stress distribution in each layer is calculated, taking into account temperature changes, phase changes and nonlinear behavior of the material.

[0031] Beneficial effects of the present invention:

[0032] The present invention provides a titanium alloy surface stress prediction method based on a subtraction average optimization algorithm, using a BP neural network optimized by the subtraction average optimization algorithm, and utilizing the high efficiency, strong adaptability, small noise impact, strong adaptability and other characteristics of the subtraction average optimization algorithm, thereby improving the accuracy and applicability of the prediction model. At the same time, both experimental data and simulation data are transmitted to the BP neural network for training, which can reduce the number of experiments, greatly reduce the time and cost of the prediction process, and more experimental sample data are used to train the neural network. Compared with the traditional method of only using experimental data to train the neural network, it can effectively improve the accuracy and efficiency of the titanium alloy surface residual stress prediction, and provide data support for the improvement of material design and processing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0034] Figure 1 The present invention is a flow chart of the titanium alloy surface stress prediction method. DETAILED DESCRIPTION

[0035] The specific implementation of the present invention is further described in detail below in conjunction with the drawings and examples.

[0036] The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0037] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections.

[0038] In order to simplify the drawings, only the parts related to the present invention are schematically shown in each figure, and they do not represent the actual structure of the product. In addition, in order to simplify the drawings and facilitate understanding, in some figures, only one of the parts with the same structure or function is schematically drawn or marked. In this article, "one" not only means "only one", but also means "more than one".

[0039] It should be further understood that the terms "and

[0040] " / or" means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0041] In the embodiments shown in the drawings, the indications of directions (such as up, down, left, right, front and back) used to explain the structure and movement of the various components of the present invention are not absolute but relative. These descriptions are appropriate when these components are in the positions shown in the drawings. If the descriptions of the positions of these components change, the indications of these directions also change accordingly.

[0042] In addition, in the description of this application, the terms "first", "second", etc. are only used to distinguish and describe.

[0043] It is not to be understood as indicating or implying relative importance.

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings and other implementation methods can be obtained based on these drawings without creative work.

[0045] See also Figure 1 The present invention provides a titanium alloy surface stress prediction method based on a subtraction average optimization algorithm, comprising the following steps:

[0046] Step 1, establishing a finite element layer-by-layer electrolysis analysis model, and obtaining surface stress simulation data through electrolysis simulation analysis;

[0047] Establish a finite element layer-by-layer electrolysis analysis model, including:

[0048] S11, model construction: Use finite element software (such as ANSYS, COMSOL, etc.) to build a three-dimensional model of the alloy, divide the alloy into multiple levels, and consider different electrolysis conditions and material properties for each layer.

[0049] S12, electrolysis process simulation: Set the chemical composition and current density of the electrolyte. Calculate the electrolysis reaction rate and thermal effect of each layer through layer-by-layer analysis.

[0050] In the above, when simulating the generation and distribution of stress of titanium alloy during processing based on the finite element method, boundary conditions and initial conditions are set. It is preferred based on the above scheme that the boundary conditions include fixed support, loading path and fine grid division; the initial conditions include current density and temperature.

[0051] Step 2, electrolyzing the titanium alloy layer by layer to obtain surface stress experimental data of each layer after electrolysis;

[0052] Specifically, the surface stress data of each layer after electrolysis treatment was measured by using X-ray diffraction method.

[0053] Step 3, verifying the electrolytic analysis model through experimental data and further optimizing the electrolytic analysis model;

[0054] Step 4, identifying and correcting the surface stress experimental data to obtain optimized surface stress experimental data;

[0055] The DBSCAN algorithm is used to identify abnormal data. The specific steps of using the DBSCAN algorithm to identify abnormal data are as follows: input the data set, randomly select an unvisited data point p from the data set to start clustering, use p as the access point, and start searching whether the Eps neighborhood of p contains Minpts data points. If so, p is marked as a core point. At the same time, the core point p creates a new cluster M, and all points in the Eps neighborhood are added to the candidate set N; for all objects q that have not been selected in the candidate set N, check whether their Eps neighborhood contains at least Minpts data points. If so, these points are added to N. If q in the candidate set N is not added to any cluster, q is added to the new cluster M, and the unprocessed data points in the candidate set N are continued to be checked. When all points in the candidate set N have been processed, stop, and repeat the above steps until all points have been visited. At this time, any points that are not added to any cluster are classified as abnormal data.

[0056] By using the DBSCAN algorithm to identify abnormal data and removing some abnormal data in the experimental data, the accuracy of the data can be improved, and ultimately the accuracy of the final results can be improved.

[0057] Step 5, using the subtraction average optimization algorithm to optimize the weights and biases of the BP neural network; the specific steps are:

[0058] S51, randomly initialize the weights and biases of the neural network;

[0059] S52, calculate the loss based on the network output and the true label;

[0060] S53, calculate the gradient of loss with respect to weights and biases;

[0061] S54, update weights and biases;

[0062] S55, repeating S52 to S54 until the loss function converges or reaches a preset number of iterations;

[0063] S56, output the results and return the optimized weights and biases.

[0064] The specific steps of updating the weights and biases in S34 are: calculating the average values ​​of the current weights and biases, and updating the weights and biases using the subtraction average method. For each weight w i and bias b j The update formula is:

[0065]

[0066] Among them, η is the learning rate, α is the weight of subtractive averaging, and w avg and b avg is the average of the current weights and biases.

[0067] The subtraction mean optimization algorithm is used to optimize the weights and biases of the BP neural network, which has the following advantages: the basic principle of the subtraction mean algorithm is relatively simple, easy to understand and implement; when processing large-scale data, the subtraction mean optimization algorithm can effectively reduce the amount of calculation, because it reduces the number of data points by selecting representative samples, which is more efficient than other algorithms. At the same time, the algorithm can adapt to different types of data distribution and is suitable for a variety of application scenarios, such as image processing, pattern recognition, etc., and has the characteristics of strong adaptability. In the process of optimization, the subtraction mean algorithm can effectively reduce the impact of noise on the results by clustering and selecting center points, thereby improving the accuracy of data analysis. The subtraction mean optimization algorithm can be used in combination with other algorithms to enhance its functionality and applicability, and has strong scalability. At the same time, the results of the subtraction mean algorithm can be visualized through the cluster center to help understand the distribution and structure of the data. And the parameters can be adjusted according to specific needs to adapt to different data sets and application scenarios.

[0068] Step 6, transmitting the optimized surface stress experimental data and surface stress simulation data to the optimized BP neural network for training, determining the number of nodes in the input layer, hidden layer and output layer, and outputting the predicted surface stress of the titanium alloy.

[0069] Sending both experimental data and simulation data to the BP neural network for training at the same time can reduce the number of experiments, significantly reduce the time and cost of the prediction process, and use more experimental sample data to train the neural network. Compared with the traditional method of using only experimental data to train the neural network, it can effectively improve the accuracy and efficiency of residual stress prediction on the titanium alloy surface.

[0070] The above are only exemplary embodiments of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and practice, those skilled in the art will easily think of other embodiments of the present disclosure. This application is intended to cover any variation, use or adaptive change of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary techniques in the technical field that are not recorded in the present disclosure.

Claims

1. A titanium alloy surface stress prediction method based on subtraction average optimization algorithm, characterized in that: The steps include: Step 1, establishing a finite element layer-by-layer electrolysis analysis model, and obtaining the surface stress simulation data of the titanium alloy by simulating and analyzing the electrolysis of the titanium alloy; Step 2, electrolyzing the titanium alloy layer by layer to obtain surface stress experimental data of each layer after electrolysis; Step 3, verifying the electrolytic analysis model through experimental data and further optimizing the electrolytic analysis model; Step 4, identifying and correcting the surface stress experimental data to obtain optimized surface stress experimental data; Step 5, using the subtraction average optimization algorithm to optimize the weights and biases of the BP neural network; Step 6, based on the optimized BP neural network, training is performed using the optimized surface stress experimental data and surface stress simulation data to predict the surface stress of the titanium alloy.

2. According to the titanium alloy surface stress prediction method based on the subtraction average optimization algorithm described in claim 1, in step 1, when simulating the titanium alloy surface stress simulation data based on the finite element method, the initial conditions and boundary conditions are set.

3. According to the titanium alloy surface stress prediction method based on subtraction average optimization algorithm in claim 2, the boundary conditions include fixed support, loading path and fine grid division; the initial conditions include current density and temperature.

4. The titanium alloy surface stress prediction method based on subtraction average optimization algorithm according to claim 1 is characterized in that: In step 3, the data set is identified and corrected, including: using the DBSCAN algorithm to identify abnormal data.

5. The titanium alloy surface stress prediction method based on subtraction average optimization algorithm according to claim 1 is characterized in that: In the step 2, the surface stress experimental data of each layer after electrolysis is obtained, specifically, the surface stress data of each layer after electrolysis is measured by using X-ray diffraction method.

6. The titanium alloy surface stress prediction method based on subtraction average optimization algorithm according to claim 1 is characterized in that: In step 5, the steps of optimizing the weights and biases of the BP neural network using the subtraction mean optimization algorithm are as follows: S51, randomly initialize the weights and biases of the neural network; S52, calculate the loss based on the network output and the true label; S53, calculate the gradient of loss with respect to weights and biases; S54, update weights and biases; S55, repeating S52 to S54 until the loss function converges or reaches a preset number of iterations; S56, output the results and return the optimized weights and biases.

7. The titanium alloy surface stress prediction method based on subtraction average optimization algorithm according to claim 6 is characterized in that: In S54, the specific steps of updating the weights and biases are: calculating the average values ​​of the current weights and biases, and updating the weights and biases using the subtraction mean method. For each weight w i and bias b j The update formula is: Among them, η is the learning rate, α is the weight of subtractive averaging, and w avg and b avg is the average of the current weights and biases.

8. The titanium alloy surface stress prediction method based on subtraction average optimization algorithm according to claim 1 is characterized in that: Establish a finite element layer-by-layer electrolysis analysis model, including: S11, model construction: 3D model of the alloy was constructed using finite element software, and the alloy was divided into multiple layers, with different electrolysis conditions and material properties considered for each layer; S12, electrolysis process simulation: set the chemical composition and current density of the electrolyte, and calculate the electrolysis reaction rate and thermal effect of each layer by layer analysis; S13, Stress Analysis: Using linear elastic theory, the stress distribution in each layer is calculated, taking into account temperature changes, phase changes and nonlinear behavior of the material.

Images