A method and device for predicting mechanical properties of TC4 titanium alloy strip

Abstract

The present invention relates to a method and device for predicting the mechanical properties of a TC4 titanium alloy coil, and belongs to the field of alloy material performance testing. Obtain the ingredients and processing parameters of the TC4 coil and the corresponding measured values ​​of the mechanical performance indicators; screen the characteristic variables of the ingredients and processing parameters, and screen out a group of characteristic variables according to the Pearson correlation coefficient and the Euclidean distance; construct a mechanical property prediction model, and optimize the weights and biases so that the loss function is less than the preset value; evaluate the mechanical property prediction model by the similarity between the test value and the measured value; process the ingredients and processing parameters of the TC4 coil that need to be predicted and input them into the mechanical property prediction model to obtain the prediction results of the mechanical performance indicators. Compared with the prior art, it saves the development cost of the TC4 coil, quickly responds to customer needs for mechanical properties, and accurately and efficiently provides ingredients and processing parameters.

Description

Technical Field

[0001] The invention relates to the field of material mechanical property testing, and in particular to a method and device for predicting the mechanical properties of a TC4 titanium alloy strip. Background Art

[0002] TC4 titanium alloy is one of the most widely used (α+β) titanium alloys at home and abroad. It has good mechanical and processing properties and is widely used in aerospace and medical fields. TC4 titanium alloy is often processed into coils and provided to subsequent processing manufacturers. TC4 titanium alloy coils are manufactured through processes such as batching, smelting, milling, rolling, heat treatment, and surface treatment, and need to achieve mechanical properties that meet customer requirements. The mechanical properties of TC4 coils are affected by many factors. For example, each element in TC4 titanium alloy affects the mechanical properties of the alloy by forming precipitation phases, refining precipitation phase size, and refining grains; heat treatment has a significant effect on the mechanical properties of titanium alloys. After quenching, the strength, toughness, and hardness of the material will change. Due to the variety of element content ratios, many process parameters, and complex mutual influence of process parameters in TC4 titanium alloy coils, it is difficult to quickly and accurately respond to customers' new requirements for mechanical properties through sample trial production, test statistics, and summarizing production experience.

[0003] In recent years, machine learning algorithms have been widely used in materials science, such as discovering new compounds, predicting physical properties, predicting crack locations, and designing steel and high-entropy alloys. However, there are still many problems with the direct application of existing machine learning algorithms to the prediction of mechanical properties of TC4 tapes: machine learning models based on composition design or process optimization are one-sided and cannot fully take into account the variables in the entire production process of TC4 tapes; TC4 tapes have many types of metal element ratios, many process parameters, and many mechanical performance indicators. The design of a combination of many components and processes is still challenging. Introducing all influencing factors will bring great computational pressure and reduce the accuracy of prediction. The existing feature screening method is single and cannot accurately screen out relevant factors in the presence of many components and process parameters; considering the large number of factors that may affect mechanical properties, the existing machine learning algorithms are not adapted to the application scenarios of TC4 tapes, and direct application will inevitably lead to inaccurate prediction accuracy and low efficiency.

[0004] In view of the above problems, designing a mechanical properties prediction method for TC4 titanium alloy strip production is an urgent problem to be solved by technical personnel in this field. Summary of the invention

[0005] The present invention proposes a method for predicting the mechanical properties of TC4 tapes, which can overcome the problems in the prior art that the variables in the entire production process of TC4 tapes cannot be fully considered, the existing feature screening method is single, and the existing algorithm is directly applied to the low accuracy of mechanical property prediction of TC4 tapes.

[0006] To achieve the above object, according to a first aspect of the present invention, a method for predicting the mechanical properties of a TC4 tape is proposed, comprising:

[0007] S1, obtaining the data of ingredients and processing parameters of TC4 coil and the corresponding measured values ​​of mechanical properties, and constructing a data set;

[0008] S2, screening the characteristic variables of ingredients and processing parameters, including normalization, screening a set of characteristic variables according to the Pearson correlation coefficient and the Euclidean distance, and dividing them into a training set and a test set;

[0009] S3, constructing a mechanical property prediction model, including an input layer, a hidden layer, and an output layer, and optimizing the weights and biases of each layer so that the loss function is less than a preset value;

[0010] S4, after inputting the test set into the mechanical property prediction model, obtaining the test value output by the mechanical property prediction model, and evaluating the mechanical property prediction model by the similarity between the test value and the measured value;

[0011] S5, the ingredients and processing parameters to be predicted are processed and input into the mechanical property prediction model to obtain the prediction results of the mechanical property indicators.

[0012] Furthermore, with respect to S1, the ingredients and processing parameters include the content of each element in the TC4 coil, smelting parameters, milling parameters, rolling parameters, and heat treatment parameters, and the mechanical performance indicators include tensile strength, yield strength, elongation after fracture, ultimate strength, elastic modulus, and specified non-proportional elongation strength.

[0013] Further, with respect to S2, the normalization process includes:

[0014] S201, use the standard Min-Max-Scalar method to map the characteristic variables to the interval [-1,1], and the calculation formula is as follows:

[0015]

[0016] In the formula, x is the original value, x in is the transformed value, x max is the maximum value of the characteristic variable, x min is the minimum value of the characteristic variable.

[0017] Furthermore, for S2, a set of characteristic variables are screened according to the Pearson correlation coefficient and the Euclidean distance, including:

[0018] S202, calculate the Pearson correlation coefficient between the characteristic variables and the measured values ​​of the mechanical performance indicators, the formula is:

[0019]

[0020] In the formula, r is the Pearson correlation coefficient, x i and i are the measured values ​​of the i-th (i=1,2,3......,N) characteristic variables and mechanical properties indicators, and are the average values ​​of the measured values ​​of characteristic variables and mechanical performance indicators, respectively, and N is the total number of samples in the data set;

[0021] S203, calculating the Euclidean distance of each characteristic variable within the data set, the formula is:

[0022]

[0023] In the formula, x i is the i-th (i=1,2,3......,N) characteristic variable, N is the total sample size of the data set;

[0024] S204, calculating the maximum value of the sum of the maximum Pearson correlation coefficient and the maximum Euclidean distance, and selecting a group of characteristic variables whose maximum value is greater than a preset threshold as the screening result, the formula is:

[0025] max(max|r|+maxED)

[0026] In the formula, max() is the maximum value function, |r| is the absolute value of the Pearson correlation coefficient, and ED is the Euclidean distance;

[0027] S205, dividing the characteristic variables in the screening results and the corresponding measured values ​​of the mechanical performance indicators into a training set and a test set in a ratio of 8:2.

[0028] Furthermore, for S3, it specifically includes:

[0029] S301, the number of nodes in the input layer is the number of characteristic variables in the screening results in step S2, the number of nodes in the hidden layer is twice the number of nodes in the input layer, and the number of nodes in the output layer is the number of mechanical performance indicators. Each layer has a corresponding weight, where W is the weight, b is the bias, and the output is expressed as:

[0030] y n =f(W n xn +b n )

[0031] In the formula, f() is the activation function, y n is the output of the nth layer, x n is the input of the nth layer, W n is the weight of the nth layer, b n is the bias of the nth layer;

[0032] S302, using the mean absolute error as the loss function of the mechanical property prediction model, representing the difference between the predicted value and the measured value, the formula is:

[0033]

[0034] Where N is the number of samples in the training set, y i is the predicted value, y i ′ is the measured value;

[0035] S303, using the training set to train the mechanical property prediction model, optimizing the weight and bias of the mechanical property prediction model, so that the loss function is less than a preset value, and the optimization formula of the weight and bias is:

[0036] m t =β1m t-1 +(1-β1) / θ1

[0037] v t =β2v t-1 +(1-β2) / θ t 2

[0038]

[0039] Where t is the number of time steps, θ is the weight or bias to be optimized, m is the first-order momentum, v is the second-order momentum, β is the hyperparameter, α is the learning rate, and ε is the optimization step to avoid The default value is zero.

[0040] Furthermore, in S4, the mechanical property prediction model is evaluated by the similarity between the test value and the measured value, including:

[0041] The R-squared value is used to evaluate the performance of the mechanical property prediction model, and the formula is:

[0042]

[0043] In the formula, N is the number of samples in the test set, y i is the predicted value, y i ′ is the measured value, is the average of N measured values.

[0044] According to a second aspect of the present invention, there is provided a device for predicting the mechanical properties of a TC4 tape, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program and implements the above-mentioned method for predicting the mechanical properties of the TC4 tape.

[0045] According to a third aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for predicting the mechanical properties of TC4 tape.

[0046] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0047] 1. Collect the data of ingredients and processing parameters during the TC4 tape processing, use all the data as the basis of the mechanical properties prediction method, comprehensively consider the factors that may affect the mechanical properties of TC4 tape, and overcome the one-sided problem of designing only from the aspects of composition or processing technology.

[0048] 2. Use the method combining Pearson correlation coefficient and Euclidean distance to screen many factors that may affect mechanical performance, so as to improve the accuracy of the prediction model, ensure the reliability of the prediction results, and reduce the computational pressure of the model.

[0049] 3. Construct a mechanical property prediction model, reasonably set the number of nodes in each layer of the model according to the number of characteristic variables and mechanical property indicators, and improve the model accuracy and calculation efficiency by optimizing the weights and biases of each layer, saving the development cost of TC4 tape, quickly responding to customer needs for mechanical properties, and accurately and efficiently providing ingredients and processing parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic diagram of a flow chart of a method for predicting mechanical properties of a TC4 tape according to an embodiment of the present invention;

[0051] Figure 2 It is a schematic diagram of the structure of the mechanical property prediction model of the present invention. DETAILED DESCRIPTION

[0052] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0053] See also Figure 1 A method for predicting mechanical properties of TC4 tape, characterized in that the steps include:

[0054] S1, obtaining the data of ingredients and processing parameters of TC4 coil and the corresponding measured values ​​of mechanical properties, and constructing a data set;

[0055] S2, screening the characteristic variables of ingredients and processing parameters, including normalization, screening a set of characteristic variables according to the Pearson correlation coefficient and Euclidean distance, and dividing them into a training set and a test set;

[0056] S3, constructing a mechanical property prediction model, including an input layer, a hidden layer, and an output layer, and optimizing the weights and biases of each layer so that the loss function is less than a preset value;

[0057] S4, after inputting the test set into the mechanical property prediction model, obtaining the test value output by the mechanical property prediction model, and evaluating the mechanical property prediction model by the similarity between the test value and the measured value;

[0058] S5, the ingredients and processing parameters to be predicted are processed and input into the mechanical property prediction model to obtain the prediction results of the mechanical property indicators.

[0059] Among them, in S1, through different channels such as experiments, sample preparation, production and processing, the process flow and ingredient parameters involved in the TC4 tape manufacturing process are determined, and the actual measured values ​​of the mechanical performance indicators corresponding to different parameters are obtained.

[0060] First, the process parameters of TC4 coil manufacturing, such as batching, smelting, milling, rolling, and heat treatment, are recorded. Specifically, the batching parameters of TC4 titanium alloy are the content of each element. The main components are Al, V, and Ti, and the impurities are Fe, C, N, H, O and other elements; the smelting parameters are arc length, arc voltage, arc current, vacuum, water temperature, etc.; the milling parameters are cutting speed, cutting depth, milling cutter diameter, etc.; the rolling parameters are rolled plate thickness, rolling temperature, rolling fire times, fire deformation, and rolling direction, etc.; the heat treatment parameters include quenching heating temperature, aging temperature, and aging time, etc.

[0061] Secondly, after each processing and manufacturing of TC4 tape, the mechanical properties of the TC4 tape are tested, and the mechanical properties indicators corresponding to different ingredients and processing parameters are recorded. Specifically, the mechanical properties indicators include tensile strength, yield strength, elongation after fracture, ultimate strength, elastic modulus, specified non-proportional extension strength, etc.

[0062] In S2, there are many characteristic variables involved in the TC4 tape processing process obtained in S1, among which there must be some characteristic variables that do not affect the mechanical properties of the TC4 tape or have a small degree of influence. If all characteristic variables are input into the mechanical properties prediction model, the accuracy of the mechanical properties prediction model will be reduced, resulting in unreliable prediction results and increasing the calculation pressure of the model. Therefore, in order to eliminate irrelevant or redundant characteristic variables, the present invention adopts a screening method based on the Pearson correlation coefficient and the Euclidean distance, specifically:

[0063] S201, use the standard Min-Max-Scalar method to map the characteristic variables to the interval [-1,1], and the calculation formula is as follows:

[0064]

[0065] In the formula, x is the original value, x in is the transformed value, x max is the maximum value of the characteristic variable, x min is the minimum value of the characteristic variable.

[0066] S202, calculate the Pearson correlation coefficient between the characteristic variables and the measured values ​​of the mechanical performance indicators, the formula is:

[0067]

[0068] In the formula, r is the Pearson correlation coefficient, x i and i are the measured values ​​of the i-th (i=1,2,3......,N) characteristic variables and mechanical properties indicators, and are the average values ​​of the characteristic variables and the measured values ​​of the mechanical performance indicators, respectively, and N is the total number of samples in the data set. i and i There is a significant correlation between them, and the absolute value of the Pearson correlation coefficient |r| will be larger.

[0069] S203, calculating the Euclidean distance of each characteristic variable within the data set, the formula is:

[0070]

[0071] In the formula, x i is the i-th (i=1,2,3......,N) characteristic variable, and N is the total sample size of the data set.

[0072] S204, calculating the maximum value of the sum of the maximum Pearson correlation coefficient and the maximum Euclidean distance, the formula is:

[0073] max(max|r|+maxED)

[0074] Where max() is the maximum value function, |r| is the absolute value of the Pearson correlation coefficient, and ED is the Euclidean distance.

[0075] Each characteristic variable has a maximum value, and a group of characteristic variables whose maximum values ​​are greater than a preset threshold are selected as screening results.

[0076] S205, the characteristic variables in the screening results and the corresponding measured values ​​of the mechanical property indicators are divided into a training set and a test set in a ratio of 8:2. The training set is used to calculate and update weights and biases, and the test set is used to evaluate the performance of the mechanical property prediction model after optimizing the parameters.

[0077] In S3, it specifically includes:

[0078] S301, construct a mechanical properties prediction model, see Figure 2 The mechanical properties prediction model is a machine learning model, which consists of an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is the number of characteristic variables in the screening results in S2, such as 8; the number of nodes in the hidden layer is twice the number of nodes in the input layer; the number of nodes in the output layer is the number of mechanical performance indicators, such as tensile strength, yield strength, ultimate strength, etc. The mechanical properties prediction model is a three-layer structure of input layer, hidden layer, and output layer. Each layer has a corresponding weight, where W is the weight and b is the bias. The output is expressed as:

[0079] y n =f(W n x n +b n )

[0080] Where f() is the activation function, which can be ELU function, ReLU function, Sigmoid function or Tanh function, y n is the output of the nth layer, x n is the input of the nth layer, W n is the weight of the nth layer, b n is the bias of the nth layer.

[0081] S302, optimizing the weight and bias of the model, wherein the mean absolute error MAE is used as the loss function of the mechanical property prediction model, indicating that the difference between the predicted value and the measured value is:

[0082]

[0083] Where N is the number of samples in the training set, y i is the predicted value, y i ′ is the measured value.

[0084] S303, using the training set to train the mechanical property prediction model, optimizing the weight and bias of the mechanical property prediction model, so that the loss function is less than a preset value, and the optimization formula of the weight and bias is:

[0085] m t =β1m t-1 +(1-β1) / θ1

[0086] v t =β2v t-1 +(1-β2) / θ t 2

[0087]

[0088] Where t is the number of time steps, θ is the weight or bias to be optimized, m is the first-order momentum, v is the second-order momentum, β is the hyperparameter, α is the learning rate, and ε is the preset value. ε is to avoid When it is zero, the denominator will be zero.

[0089] In S4, after the mechanical property prediction model is constructed, the data in the test set is input into the mechanical property prediction model, the test value output by the mechanical property prediction model is obtained, the predicted result is compared with the actual result, and the mechanical property prediction model is evaluated by the similarity between the test value and the measured value.

[0090] The R-squared value is used to evaluate the model performance. The R-squared value indicates the degree of fit between the predicted result and the actual result. The closer the R-squared value is to 1, the better the model fit is. The R-squared value calculation formula is as follows:

[0091]

[0092] In the formula, N is the number of samples in the test set, y i is the predicted value, y i ′ is the measured value, is the average of N measured values.

[0093] In S5, the TC4 tape ingredients and process parameters to be predicted are processed and input into the mechanical property prediction model to predict its mechanical property indicators.

[0094] The present invention provides a device for predicting the mechanical properties of a TC4 tape, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program and implements the method for predicting the mechanical properties of the TC4 tape as described in the aforementioned embodiment.

[0095] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a microprocessor, the method for predicting the mechanical properties of the TC4 tape described in the above embodiment is implemented.

[0096] It is worth noting that the system, electronic device and computer-readable storage medium described in the present invention are all based on the same inventive concept as the method for predicting the mechanical properties of a TC4 tape in Example 1, and the specific technical content will not be repeated here.

[0097] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting the mechanical properties of TC4 tape, characterized in that the steps include: S1, obtaining the data of ingredients and processing parameters of TC4 coil and the corresponding measured values ​​of mechanical properties, and constructing a data set; S2, screening the characteristic variables of ingredients and processing parameters, including normalization processing, calculating the Pearson correlation coefficient between the characteristic variables and the measured values ​​of the mechanical performance indicators and the Euclidean distance of each characteristic variable within the data set, calculating the maximum value of the sum of the maximum Pearson correlation coefficient and the maximum Euclidean distance, selecting a group of characteristic variables whose maximum value is greater than a preset threshold as the screening result, and dividing them into a training set and a test set; S3, construct a mechanical property prediction model, including an input layer, a hidden layer, and an output layer. By optimizing the weights and biases of each layer, the loss function is made less than the preset value. The optimization formulas for weights and biases are: m t =β1m t-1 +(1-β1) / θ1 v t =β2v t-1 +(1-β2) / θ t 2 Where t is the number of time steps, θ is the weight or bias to be optimized, m is the first-order momentum, v is the second-order momentum, β is the hyperparameter, α is the learning rate, and ε is the optimization step to avoid The default value is zero; S4, after inputting the test set into the mechanical property prediction model, obtaining the test value output by the mechanical property prediction model, and evaluating the mechanical property prediction model by the similarity between the test value and the measured value; S5, processing the ingredients and processing parameters to be predicted and inputting them into the mechanical properties prediction model to obtain the prediction results of the mechanical properties index; Among them, in S1, the ingredients and processing parameters include the content of each element in the TC4 coil, smelting parameters, milling parameters, rolling parameters, and heat treatment parameters, and the mechanical performance indicators include tensile strength, yield strength, elongation after fracture, ultimate strength, elastic modulus, and specified non-proportional extension strength; In S2, a set of characteristic variables are screened out according to the Pearson correlation coefficient and Euclidean distance, including: S202, calculate the Pearson correlation coefficient between the characteristic variables and the measured values ​​of the mechanical performance indicators, the formula is: In the formula, r is the Pearson correlation coefficient, x i and i ′ are the measured values ​​of the i-th (i=1,2,3......,N) characteristic variables and mechanical properties indicators, and are the average values ​​of the measured values ​​of characteristic variables and mechanical performance indicators, respectively, and N is the total number of samples in the data set; S203, calculating the Euclidean distance of each characteristic variable within the data set; S204, select a group of characteristic variables with the largest sum of the maximum Pearson correlation coefficient and the maximum Euclidean distance as the screening result, the formula is: max(max|r|+maxED) In the formula, max() is the maximum value function, |r| is the absolute value of the Pearson correlation coefficient, and ED is the Euclidean distance; S205, dividing the characteristic variables in the screening results and the corresponding measured values ​​of the mechanical performance indicators into a training set and a test set in a ratio of 8:2; In S3, it specifically includes: S301, the number of nodes in the input layer is the number of characteristic variables in the screening results in step S2, the number of nodes in the hidden layer is twice the number of nodes in the input layer, and the number of nodes in the output layer is the number of mechanical performance indicators. Each layer has a corresponding weight, where W is the weight, b is the bias, and the output is expressed as: y n =f(W n x n +b n ) In the formula, f() is the activation function, y n is the output of the nth layer, x n is the input of the nth layer, W n is the weight of the nth layer, b n is the bias of the nth layer; S302, using the mean absolute error as the loss function of the mechanical property prediction model, representing the difference between the predicted value and the measured value, the formula is: Where N is the number of samples in the training set, y i is the predicted value, y i ' is the measured value; The training set is used to train the mechanical property prediction model, and the weights and biases of the mechanical property prediction model are optimized so that the loss function is less than the preset value.

2. A TC4 tape mechanical property prediction method according to claim 1, characterized in that: In S2, the normalization process includes: S201, use the standard Min-Max-Scalar method to map the characteristic variables to the interval [-1,1], and the calculation formula is as follows: In the formula, x is the original value, x in is the transformed value, x max is the maximum value of the characteristic variable, x min is the minimum value of the feature variable.

3. The method for predicting mechanical properties of TC4 tape according to claim 1, characterized in that: In S4, the mechanical property prediction model is evaluated by the similarity between the test value and the measured value, including: using the R square value to evaluate the performance of the mechanical property prediction model, the formula is: In the formula, N is the number of samples in the test set, y i is the predicted value, y i ′ is the measured value, is the average of the measured values ​​of N samples.

4. A device for predicting the mechanical properties of a TC4 tape, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program and implements the method for predicting the mechanical properties of a TC4 tape as claimed in any one of claims 1 to 3.

5. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the method for predicting the mechanical properties of a TC4 tape as claimed in any one of claims 1 to 3 is implemented.

Images