Method for predicting tensile strength of additive manufacturing 316L stainless steel based on machine learning

Abstract

The invention discloses a method for predicting the tensile strength of additive manufacturing 316L stainless steel based on machine learning. The method comprises the following steps: firstly, establishing a machine learning data set which takes four process parameters of laser power, scanning speed, scanning spacing and powder laying layer thickness as input and takes the tensile strength of SLM (Selective Laser Melting) molding 316L stainless steel as output; then preprocessing and normalizing data in the machine learning data set; constructing and training a tensile strength prediction model composed of a convolutional neural network and a random forest regression device; and finally, obtaining a tensile strength prediction result by using the trained tensile strength prediction model. According to the method, high-precision prediction of small data set samples can be realized.

Description

Technical Field

[0001] The present invention relates to the field of additive manufacturing mechanical property prediction, and more specifically to a method for predicting the tensile strength of 316L stainless steel in additive manufacturing based on machine learning. Background Art

[0002] As an important branch of additive manufacturing (AM) technology, selective laser melting (SLM) has demonstrated its revolutionary manufacturing potential in aerospace, automotive industry, mold manufacturing, hydraulic components, medical devices and other fields in recent years. In the molding process, a high-energy laser is used to scan each slice according to a planned path. The powder scanned by the laser spot is quickly melted and solidified, and after accumulation between layers, the three-dimensional model is finally formed. Compared with traditional manufacturing technologies such as casting, forging, and machining, SLM has the advantages of fast molding speed, high precision, high material utilization rate, and integrated molding. The four process parameters of laser power, scanning speed, scanning spacing, and powder layer thickness largely determine the mechanical properties of SLM-molded metal parts. The influence of process parameters on mechanical properties is obtained through experiments, which will consume a lot of time and manpower costs. Therefore, it is urgent to establish a method for predicting the mechanical properties of SLM molding to quickly and efficiently achieve the regulation of mechanical properties. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention proposes a method for predicting the tensile strength of 316L stainless steel in additive manufacturing based on machine learning. The specific technical solution is as follows:

[0004] A method for predicting the tensile strength of 316L stainless steel in additive manufacturing based on machine learning, the method comprising the following steps:

[0005] Step 1: Establish a machine learning dataset with four process parameters, laser power, scanning speed, scanning spacing, and powder layer thickness, as input and the tensile strength of 316L stainless steel formed by SLM as output;

[0006] Step 2: preprocessing and normalizing the data in the machine learning dataset;

[0007] Step 3: Construct and train a tensile strength prediction model; the tensile strength prediction model consists of a convolutional neural network and a random forest regressor; the convolutional neural network includes three convolutional layers and two fully connected layers, the convolutional layer is used to perform a convolution operation on the input, extract features, and then generate output through the ReLU activation function, flattening and the fully connected layer; the random forest regressor is used to perform further regression analysis on the output generated by the fully connected layer;

[0008] Step 4: After the laser power, scanning speed, scanning spacing, and powder layer thickness of the additively manufactured 316L stainless steel to be predicted are preprocessed and normalized in the same way as in step 2, they are input into the tensile strength prediction model trained in step 3 to obtain the tensile strength prediction result.

[0009] Furthermore, in step 1, the laser power is between 100 and 300 W, and the scanning speed is between 600 and 1000 mm·s -1 The scanning interval is between 0.1 and 0.14 mm, and the powder layer thickness is between 30 and 50 μm; the tensile strength of 316L stainless steel formed under different process parameters is measured by tensile test.

[0010] Furthermore, in the step 2, the data in the machine learning dataset is normalized by MinMaxScaler and converted into a PyTorch tensor.

[0011] Furthermore, in the step three, when the convolutional neural network is trained, the model is trained for 50 generations using the MSE loss function.

[0012] Furthermore, the eigenvector of the random forest regressor is 17, the minimum number of samples is 3, and the number of trees in the random forest is 81.

[0013] The beneficial effects of the present invention are as follows:

[0014] The machine learning-based additively manufactured 316L stainless steel tensile strength prediction method of the present invention can realize the prediction of the mechanical properties of additively manufactured 316L stainless steel with a small data set sample, and the prediction has high accuracy and strong reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Flowchart of the tensile strength prediction method for additively manufactured 316L stainless steel based on machine learning.

[0016] Figure 2 It is a comparison chart of the predicted values ​​and experimental values ​​of the embodiments of the present invention and the comparative examples.

[0017] Figure 3 It is a comparison chart of evaluation indicators of the embodiments of the present invention and the comparative examples. DETAILED DESCRIPTION

[0018] The present invention will be described in detail below based on the accompanying drawings and preferred embodiments, and the purpose and effects of the present invention will become more clear. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] Based on the established machine learning data set, the present invention reads, cleans and normalizes the data, then defines a convolutional neural network (CNN) model, trains it, and evaluates the performance of the model on the validation set. Finally, the output of the CNN model is further regressed through a random forest (RF) regressor, and various evaluation indicators are calculated. The schematic diagram of the prediction training process is shown in Figure 1 shown.

[0020] like Figure 1 As shown, the method for predicting the tensile strength of 316L stainless steel in additive manufacturing based on machine learning in an embodiment of the present invention specifically includes the following steps:

[0021] Step 1: Establish a machine learning dataset with four process parameters as input: laser power, scanning speed, scanning spacing, and powder layer thickness, and the tensile strength of 316L stainless steel formed by SLM as output. The laser power is between 100 and 300 W, and the scanning speed is between 600 and 1000 mm·s -1 The scanning interval is between 0.1 and 0.14 mm, and the powder layer thickness is between 30 and 50 μm.

[0022] This embodiment designs an orthogonal experiment and measures the tensile strength of 316L stainless steel formed under different process parameters through a tensile test, and obtains the following Table 1.

[0023] Table 1 Tensile strength of 316L stainless steel formed by SLM

[0024]

[0025]

[0026] Step 2: Preprocess the data in step 1, and then divide the data into features and target values; the features and target values ​​are normalized by MinMaxScaler and converted into PyTorch tensors.

[0027] Step 3: Build and train the tensile strength prediction model. The tensile strength prediction model consists of a convolutional neural network and a random forest regressor. The convolutional neural network (CNN) consists of three convolutional layers and two fully connected layers, which is suitable for processing data with a specific spatial structure. The convolution layer performs a convolution operation on the input to extract features, and then uses the ReLU activation function to flatten and generate output through the fully connected layer. The fully connected layer maps these features to the final output. The model was trained for 50 generations using the MSE loss function, and the model was evaluated using the Adam optimizer to calculate the relevant evaluation criteria mean square error (MSE), mean absolute error (MAE), and correlation coefficient R 2MSE is used to measure the average degree of the square of the difference between the predicted value and the actual value. The smaller its value, the smaller the prediction error of the model and the better the prediction performance. MAE reflects the average size of the difference between the predicted value and the actual value. The smaller the value, the higher the prediction accuracy. 2 The statistical index used to characterize the goodness of fit of the regression model, R 2 The closer it is to 1, the better the model fits the data. MSE, MAE, R 2 Calculated by formula (1)-(3):

[0028]

[0029]

[0030]

[0031] Where n is the total number of data points, X i is the actual value, Y i is the predicted value, is the average of the predicted values.

[0032] Finally, the CNN output is input as a feature into the Random Forest (RF) regression model, a Random Forest Regressor instance is created, and it is trained using the training data.

[0033] In this embodiment, the number of eigenvectors considered for each split is 17 through cross validation. The minimum number of samples for each internal node is 3, and too small a sample may lead to overfitting. The number of trees in the random forest is 81. Increasing the number of trees usually improves the performance of the model, but it also increases the computational cost. Further regression analysis is performed through the random forest regressor to complete the calculation of predictions and various evaluation indicators.

[0034] This example verifies the accuracy of model prediction through 12 sets of experiments. Figure 2 The red and blue points in the figure represent the overlap between the predicted value and the experimental value when using CNN alone and CNN+RF for prediction, respectively. It can be seen that the distribution of CNN+RF points is closer to the diagonal line of the frame, indicating that the predicted value is closer to the experimental value, indicating that the embodiment of the present invention significantly improves the accuracy of model prediction. Figure 3 The comparison of the prediction evaluation indicators of CNN and CNN+RF shows that compared with CNN, the MSE and MAE of CNN+RF dropped from 0.00305 and 0.0505 to 0.00295 and 0.0344, respectively, down by 3.28% and 31.88%, while R 2It increased from 0.2473 to 0.9576, an increase of 74.18%, indicating that the reliability of the predicted value has been significantly enhanced. The CNN+RF model can better fit complex nonlinear problems and obtain more accurate prediction results when using small data set samples for prediction.

[0035] Those skilled in the art can understand that the above are only preferred examples of the invention and are not intended to limit the invention. Although the invention is described in detail with reference to the above examples, those skilled in the art can still modify the technical solutions recorded in the above examples or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, etc. made within the spirit and principle of the invention shall be included in the protection scope of the invention.

Claims

1. A method for predicting the tensile strength of 316L stainless steel in additive manufacturing based on machine learning, characterized in that: The method comprises the following steps: Step 1: Establish a machine learning dataset with four process parameters as input: laser power, scanning speed, scanning spacing, and powder layer thickness, and the tensile strength of 316L stainless steel formed by SLM as output; Step 2: preprocessing and normalizing the data in the machine learning dataset; Step 3: Construct and train a tensile strength prediction model; the tensile strength prediction model consists of a convolutional neural network and a random forest regressor; the convolutional neural network includes three convolutional layers and two fully connected layers, the convolutional layer is used to perform a convolution operation on the input, extract features, and then generate output through the ReLU activation function, flattening and the fully connected layer; the random forest regressor is used to perform further regression analysis on the output generated by the fully connected layer; Step 4: After the laser power, scanning speed, scanning spacing, and powder layer thickness of the additively manufactured 316L stainless steel to be predicted are preprocessed and normalized in the same way as in step 2, they are input into the tensile strength prediction model trained in step 3 to obtain the tensile strength prediction result.

2. The method for predicting tensile strength of 316L stainless steel in additive manufacturing based on machine learning according to claim 1, characterized in that: In step 1, the laser power is between 100 and 300 W, and the scanning speed is between 600 and 1000 mm·s. -1 The scanning interval is between 0.1 and 0.14 mm, and the powder layer thickness is between 30 and 50 μm; the tensile strength of 316L stainless steel formed under different process parameters is measured by tensile test.

3. The method for predicting tensile strength of 316L stainless steel in additive manufacturing based on machine learning according to claim 1, characterized in that: In the step 2, the data in the machine learning dataset is normalized by MinMaxScaler and converted into a PyTorch tensor.

4. The method for predicting tensile strength of 316L stainless steel in additive manufacturing based on machine learning according to claim 1, characterized in that: In the step three, when the convolutional neural network is trained, the model is trained for 50 generations using the MSE loss function.

5. The method for predicting tensile strength of 316L stainless steel in additive manufacturing based on machine learning according to claim 1, characterized in that: The number of eigenvectors for the random forest regressor is 17, the minimum number of samples is 3, and the number of trees in the random forest is 81.

Images