A Sensitivity Analysis Method for Factors Affecting Deformation in Arc Additive Manufacturing

Abstract

This invention discloses a sensitivity analysis method for deformation in arc additive manufacturing, comprising: determining the research object and the influencing factors of deformation in arc additive manufacturing; verifying a pre-constructed thermo-mechanical coupling numerical calculation model and generating a dataset through the thermo-mechanical coupling numerical calculation model; constructing a machine learning model, including an artificial neural network model and a random forest model, and training the machine learning model using the dataset; predicting component deformation using the machine learning model; drawing deformation contour maps of process parameters and analyzing the sensitivity of influencing factors, and analyzing the sensitivity of influencing factors based on the feature importance values ​​of parameters in the random forest model; comparing the obtained sensitivity ranking of influencing factors to verify the accuracy of the proposed analysis method. This invention solves the problems of complex mechanisms in numerical simulation methods and lack of training datasets in machine learning methods, and addresses the issue that the influence of a single factor on deformation is easily masked by the complexity of the combined effects of multiple factors.

Description

Technical Field

[0001] This invention pertains to metal additive manufacturing technology, specifically relating to a sensitivity analysis method for factors influencing deformation in arc additive manufacturing. Background Technology

[0002] In recent years, arc additive manufacturing technology has attracted widespread attention in the manufacturing industry due to its advantages such as high deposition rate, variety of wire materials, high material utilization, and low cost, and is considered a promising alternative technology for manufacturing large metal parts. This technology is relatively mature abroad and has been applied in high-end equipment manufacturing fields such as aerospace. However, there is still no reasonable and effective method to rank the sensitivity of factors affecting deformation in arc additive manufacturing. During the arc additive manufacturing process, the deformation of the component is affected by multiple factors, such as preheating temperature, arc power, melt flow overlap rate, and scanning speed. The process evolution and mechanism are complex, and the influence of a single factor on deformation is easily masked by the complexity of the combined effects of multiple factors, making it difficult to analyze the sensitivity of factors affecting deformation in arc additive manufacturing using traditional numerical simulation methods.

[0003] Investigating the sensitivity of various influencing factors on deformation in arc additive manufacturing using traditional methods inevitably requires extensive experiments and complex analytical approaches, resulting in significant labor and material costs, as well as considerable time and effort. Compared to numerical simulation methods, machine learning can simplify the complex relationship between input and output, providing an effective means to study the impact of various factors on deformation. Therefore, it is necessary to design an efficient and reasonable sensitivity analysis method that incorporates machine learning. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide a sensitivity analysis method for the influencing factors of deformation in arc additive manufacturing, which reasonably determines the sensitivity ranking of the influencing factors of deformation in arc additive manufacturing, and solves the problem that the influence of a single factor on deformation is easily covered by the complexity of the combined effect of multiple factors.

[0005] Technical Solution: This invention provides a sensitivity analysis method for deformation in arc additive manufacturing, comprising: determining the research object and the influencing factors of deformation in arc additive manufacturing; verifying a pre-constructed thermo-mechanical coupling numerical calculation model and generating a dataset using the verified thermo-mechanical coupling numerical calculation model; constructing a machine learning model, which includes an artificial neural network model and a random forest model, and training the machine learning model using the generated dataset; predicting the maximum deformation of thin-walled parts manufactured by arc additive manufacturing using the trained machine learning model; drawing deformation contour map of process parameters and analyzing the sensitivity of influencing factors, while simultaneously analyzing the sensitivity of influencing factors based on the feature importance values ​​of parameters in the random forest model; and comparing the sensitivity ranking of influencing factors obtained through the deformation contour map of process parameters with the sensitivity ranking of influencing factors obtained based on the feature importance values ​​of parameters in the random forest machine learning model to verify the accuracy of the proposed parameter sensitivity comparison method.

[0006] The pre-constructed thermo-mechanical coupling numerical calculation model is validated by: selecting a set of typical parameters affecting deformation in arc additive manufacturing, establishing a thermo-mechanical coupling numerical calculation model under these parameters, and calculating the maximum deformation of the thin-walled part under these parameters; conducting physical experiments using these process parameters to measure the maximum deformation of the printed thin-walled part corresponding to different values ​​of the process parameters; comparing the maximum deformation of the thin-walled part calculated by the thermo-mechanical coupling numerical calculation model with the actual measured value of the maximum deformation of the thin-walled part in the physical experiment to determine whether they are consistent, thereby validating the thermo-mechanical coupling numerical calculation model and verifying its accuracy in order to obtain the correct dataset for training the machine learning model.

[0007] The dataset is generated by the validated thermo-mechanical coupling numerical calculation model, including: generating the dataset by using the full factorial experimental method with four factors and four levels to avoid missing the influence of a certain process parameter on the deformation of arc additive manufacturing.

[0008] The artificial neural network model includes the following hyperparameters: number of network layers, number of hidden layer neurons, learning rate, and number of learning epochs. The number of hidden layers, number of hidden layer neurons, and activation function are selected as key optimization factors. A random search method is used to adjust the parameters and find the optimal hyperparameters. An optimal artificial neural network model is established to better predict the deformation in arc additive manufacturing.

[0009] The generated dataset is used to train a machine learning model, including: selecting mean absolute error and relative coefficient as evaluation indicators; mean absolute error is the average absolute error between the deformation predicted by the machine learning model and the deformation calculated by numerical simulation; the relative coefficient is used to quantify the deviation between the data and the mean, and the relative coefficient ranges from [0,1]. When the relative coefficient is 0, it indicates that the model fits poorly, and when the relative coefficient is 1, it indicates that the deformation predicted by the model is consistent with the deformation calculated by numerical simulation; to ensure that the trained machine learning model can accurately predict the deformation of arc additive manufacturing based on the input process parameters.

[0010] Plotting deformation contour maps of process parameters and analyzing the sensitivity of influencing factors, including: using the minimax normalization method to normalize the process parameters on both the horizontal and vertical axes when plotting the contour maps; the calculation formula is as follows. V represents the original variable value, V min and V max V represents the minimum and maximum values ​​of variable V, respectively. V* represents the normalized variable, with the value range of the normalized variable being [0,1]. The horizontal and vertical coordinate ranges of the deformation contour line process parameter map are unified to facilitate comparison of the sensitivity of the two process parameters.

[0011] The process of plotting deformation contour maps and analyzing the sensitivity of influencing factors also includes: analyzing the slope of the deformation contour lines in the process parameter maps for arc additive manufacturing to assess the degree of influence of the process parameters on the deformation of arc additive manufacturing on the horizontal and vertical axes. This degree of influence is the sensitivity of arc additive manufacturing deformation to process parameters. By comparing the slopes of the contour lines in the normalized deformation contour maps, the sensitivity of the process parameters represented by the horizontal and vertical axes can be compared simply and intuitively.

[0012] When the slope of the arc additive manufacturing deformation contour line is less than the preset threshold, it indicates that the process parameters corresponding to the vertical axis have a significant impact on the deformation of the arc additive manufacturing component; conversely, when the slope of the arc additive manufacturing deformation contour line is greater than the preset threshold, it indicates that the process parameters corresponding to the horizontal axis have a significant impact on the deformation of the arc additive manufacturing component.

[0013] The present invention also includes an apparatus comprising a memory and a processor, wherein:

[0014] Memory is used to store computer programs that can run on a processor;

[0015] A processor is configured to execute, while running the computer program, the steps of the sensitivity analysis method for the influencing factors of deformation in arc additive manufacturing.

[0016] The present invention also includes a computer-readable storage medium storing a computer program that, when executed by at least one processor, implements the steps of the sensitivity analysis method for deformation in arc additive manufacturing.

[0017] Beneficial Effects: Compared with existing technologies, the technical solution of this invention has the following beneficial effects: By combining numerical simulation and machine learning methods, it solves the problems of complex mechanisms in numerical simulation methods and lack of training datasets in machine learning methods; by comparing the slope of contour lines in the deformation contour map to determine the sensitivity of process parameters, it solves the problem that the influence of a single factor on deformation is easily obscured by the complexity of the combined effects of multiple factors. This invention greatly saves manpower and resources, simplifies the relationship between various process parameters and deformation in arc additive manufacturing, and scientifically and rationally determines the sensitivity ranking of influencing factors on deformation in arc additive manufacturing. Attached Figure Description

[0018] Figure 1 This is a flowchart of the present invention;

[0019] Figure 2 This is a schematic diagram of the structure of a double-track ten-layer forming part model according to an embodiment of the present invention;

[0020] Figure 3 This is a schematic diagram of the finite element simulation model of an embodiment of the present invention;

[0021] Figure 4 for Figure 3 The left view;

[0022] Figure 5 This is a schematic diagram of the artificial neural network structure according to an embodiment of the present invention;

[0023] Figure 6 This is a schematic diagram of a random forest structure according to an embodiment of the present invention;

[0024] Figure 7 This is a process parameter diagram of deformation contour lines according to an embodiment of the present invention;

[0025] Figure 8 This is a normalized deformation contour line process parameter diagram according to an embodiment of the present invention. Detailed Implementation

[0026] The technical solution of the present invention will now be described in detail with reference to specific embodiments and accompanying drawings.

[0027] like Figure 1 As shown, the present invention provides a sensitivity analysis method for factors affecting deformation in arc additive manufacturing, comprising the following steps:

[0028] S1: Determine the research object and the influencing factors of deformation in arc additive manufacturing.

[0029] The research object is the component manufactured by electric arc additive manufacturing. In this scheme, the maximum deformation of thin-walled parts is selected as the test index. The influencing factors include four key process parameters that affect the deformation of electric arc additive manufacturing, namely, preheating temperature, arc power, scanning speed, and melt overlap rate. The values ​​of the four process parameters are divided into four levels, namely "high, relatively high, medium, and low". The values ​​of the process parameters are shown in Table 1.

[0030] In this embodiment, the wire used is ER5356 with a diameter of 1.2mm, and the substrate is 6061 aluminum alloy with a diameter of 240mm×240mm×10mm.

[0031] Table of process parameter values

[0032]

[0033] S2: Construct a thermo-mechanical coupling numerical calculation model and verify its accuracy.

[0034] like Figure 2 and Figure 3 As shown in this embodiment, a dual-channel, ten-layer thermo-coupling numerical calculation model is established using the finite element simulation method.

[0035] A set of typical parameters affecting deformation in arc additive manufacturing was selected, and a thermo-mechanical coupling numerical calculation model was established under these parameters to calculate the maximum deformation of thin-walled parts. Physical experiments were conducted using these process parameters, measuring the maximum deformation of the printed thin-walled parts corresponding to different parameter values. The maximum deformation of the thin-walled parts calculated by the thermo-mechanical coupling numerical calculation model was compared with the actual measured values ​​in the physical experiments to verify the model. The comparison revealed that the maximum deformation of the thin-walled parts calculated by the thermo-mechanical coupling numerical calculation model was basically consistent with the actual measured values, indicating that the calculation results of the thermo-mechanical coupling numerical calculation model have high accuracy.

[0036] S3: The dataset is generated using the full factorial experimental method through the validated thermo-mechanical coupling numerical calculation model. The full factorial experimental method uses four factors and four levels.

[0037] S4: As Figure 4 and Figure 5 As shown, a machine learning model is established, which includes an artificial neural network model and a random forest model.

[0038] Artificial neural network models include hyperparameters such as the number of network layers, the number of hidden layer neurons, the learning rate, and the number of learning epochs. To simplify the model parameter tuning process, the number of hidden layers, the number of hidden layer neurons, and the activation function are selected as key optimization factors. A random search method is used to adjust the parameters and find the optimal hyperparameters.

[0039] S5: Use the generated dataset to train a machine learning model.

[0040] 70% of the numerical simulation results were randomly selected as the training dataset, 20% as the validation dataset, and the remaining 10% as the test dataset. Artificial neural network models and random forest models were trained, validated, and tested respectively. The mean absolute error (MAE) and relative coefficient (R²) were used as evaluation metrics for training the machine learning models. 2 value).

[0041] MAE indicates the mean absolute error between the deformation predicted by the machine learning model and the deformation calculated by numerical simulation. A smaller MAE indicates that the machine learning model has higher accuracy; R 2 The value quantifies the deviation of the data from the mean, and its range is [0,1]. When R... 2 A value of 0 indicates a poor model fit; when R0 is 0, the model fit is poor. 2 A value of 1 indicates that the deformation predicted by the model is in complete agreement with the deformation calculated numerically.

[0042] S6: Utilize a trained machine learning model to predict the maximum deformation of thin-walled parts manufactured by arc additive manufacturing.

[0043] To obtain the sensitivity of various factors to deformation in arc additive manufacturing, pairwise comparisons are required for the four selected typical process parameters. Therefore, keeping two process parameters constant while changing the other two, these parameters are used as input to a trained machine learning model to accurately predict component deformation. This pairwise combination needs to be repeated six times. For example, to compare the sensitivity of preheating temperature and arc power to deformation, the scanning speed and melt flow overlap are kept constant, while the values ​​of preheating temperature and arc power are changed. These values ​​are then input into a trained artificial neural network model and a random forest model to predict the maximum deformation of thin-walled parts manufactured by arc additive manufacturing.

[0044] S7: As Figure 6 As shown, deformation contour maps of process parameters are plotted and the sensitivity of influencing factors is analyzed. At the same time, the sensitivity of influencing factors is analyzed based on the characteristic importance values ​​of parameters in the random forest model.

[0045] The factors influencing deformation in arc additive manufacturing listed in this example include: preheating temperature, arc power, scanning speed, and weld overlap rate. Since pairwise comparisons of these four parameters are required six times, six deformation contour plots are drawn. Taking the deformation contour plot of preheating temperature and arc power as an example: comparing the sensitivity of preheating temperature and arc power to deformation in arc additive manufacturing, and using the controlled variable method, keeping the values ​​of scanning speed and weld overlap rate constant, the preheating temperature and arc power are varied within the ranges of 20-320 (°C) and 750-2000 (W), respectively. Based on the maximum deformation predicted in step S6, the deformation contour plots of preheating temperature and arc power are drawn. The principle for drawing other deformation contour plots is the same.

[0046] The contour plot of process parameters uses the minimax normalization method to normalize the process parameters on both the horizontal and vertical axes. The calculation formula is as follows: V is the original variable value, V min and V max V represents the minimum and maximum values ​​of variable V, respectively. V* is the normalized variable, and its value range is [0,1].

[0047] The normalized deformation contour map of process parameters allows for the assessment of the influence (i.e., sensitivity) of the process parameters on the deformation of arc additive manufacturing components by analyzing the slope of the contour lines. When the slope of the deformation contour lines is less than 1, it indicates a greater influence (i.e., higher sensitivity) of the process parameters corresponding to the vertical axis on the deformation of the arc additive manufacturing components; conversely, when the slope is greater than 1, it indicates a greater influence (i.e., higher sensitivity) of the process parameters corresponding to the horizontal axis on the deformation of the components. Figure 7 As shown.

[0048] Plot the deformation contour lines and process parameters, and analyze the sensitivity of influencing factors. The original deformation contour lines and process parameters are shown below. Figure 6 As shown, the normalized deformation contour map shows the process parameters. Figure 7 As shown, the order of influence is: preheating temperature > arc power > melt path overlap rate > scanning speed. At the same time, the sensitivity of influencing factors is analyzed based on the feature importance values ​​of the parameters in the random forest model. The feature importance values ​​are inherent parameters of the random forest model.

[0049] S8: Verify the accuracy of the proposed parameter sensitivity comparison method.

[0050] The sensitivity ranking of influencing factors obtained from the deformation contour map of process parameters is compared with the sensitivity ranking of influencing factors obtained from the feature importance values ​​of parameters in random forest machine learning to verify the accuracy of the proposed parameter sensitivity comparison method. The sensitivity ranking of process parameters obtained by this scheme is the same as that obtained by the feature importance values ​​of random forest, with preheating temperature > arc power > melt channel overlap rate > scanning speed.

[0051] The present invention also includes an apparatus comprising a memory and a processor, wherein:

[0052] Memory is used to store computer programs that can run on a processor;

[0053] A processor, used to perform the steps of a sensitivity analysis method for factors affecting deformation in arc additive manufacturing while running the computer program.

[0054] The present invention also includes a computer-readable storage medium storing a computer program that, when executed by at least one processor, implements the steps of the sensitivity analysis method for the influence factors of deformation in arc additive manufacturing.

Claims

1. A sensitivity analysis method for factors affecting deformation in arc additive manufacturing, characterized in that, include: Identify the research subjects and the influencing factors of deformation in arc additive manufacturing; The pre-built thermo-coupling numerical calculation model is validated, and a dataset is generated using the validated thermo-coupling numerical calculation model. Construct a machine learning model, which includes an artificial neural network model and a random forest model, and train the machine learning model using the generated dataset; To obtain the sensitivity of each influencing factor to the deformation of arc additive manufacturing, pairwise comparisons are required for the four selected typical process parameters, and a trained machine learning model is used to predict the maximum deformation of thin-walled parts manufactured by arc additive manufacturing. Plot deformation contour lines and process parameter diagrams and analyze the sensitivity of influencing factors. At the same time, analyze the sensitivity of influencing factors based on the characteristic importance values ​​of parameters in the random forest model. The sensitivity ranking of influencing factors obtained by the process parameter map of deformation contour lines is compared with the sensitivity ranking of influencing factors obtained by the feature importance value of parameters in random forest machine learning to verify the accuracy of the proposed parameter sensitivity analysis method. The pre-built thermo-coupling numerical calculation model was validated, including: A set of typical parameters affecting the deformation of arc additive manufacturing is selected, and a thermo-mechanical coupling numerical calculation model under this set of parameters is established to solve the maximum deformation of thin-walled parts under this set of parameters. Physical experiments were conducted using this set of process parameters to measure the maximum deformation of the printed thin-walled parts corresponding to different values ​​of the process parameters. The maximum deformation of the thin-walled component calculated by the thermo-mechanical coupling numerical calculation model is compared with the actual measured value of the maximum deformation of the thin-walled component in the physical experiment to determine whether they are consistent, thereby verifying the thermo-mechanical coupling numerical calculation model.

2. The sensitivity analysis method for deformation in arc additive manufacturing according to claim 1, characterized in that, A dataset was generated using the validated thermo-coupling numerical computation model, including: A dataset was generated using a full factorial experimental method with four factors and four levels, based on a validated thermo-mechanical coupling numerical calculation model.

3. The sensitivity analysis method for deformation in arc additive manufacturing according to claim 1, characterized in that, The artificial neural network model includes the following hyperparameters: number of network layers, number of hidden layer neurons, learning rate, and number of learning epochs. The number of hidden layers, number of hidden layer neurons, and activation function are selected as key optimization factors. A random search method is used to adjust the parameters and find the optimal hyperparameters.

4. The sensitivity analysis method for deformation in arc additive manufacturing according to claim 1, characterized in that, The generated dataset is used to train a machine learning model, including: selecting mean absolute error and relative coefficient as evaluation indicators; mean absolute error is the average absolute error between the deformation predicted by the machine learning model and the deformation calculated by numerical simulation; the relative coefficient is used to quantify the deviation between the data and the mean, and the relative coefficient ranges from [0,1]. When the relative coefficient is 0, it indicates that the model fits poorly, and when the relative coefficient is 1, it indicates that the deformation predicted by the model is consistent with the deformation calculated by numerical simulation.

5. The sensitivity analysis method for deformation in arc additive manufacturing according to claim 1, characterized in that, Plot deformation contour lines and process parameter diagrams, and analyze the sensitivity of influencing factors, including: When plotting contour lines and process parameters, the minimax normalization method is used to normalize the process parameters on both the horizontal and vertical axes. The calculation formula is as follows: V represents the original variable value, V min and V max V represents the minimum and maximum values ​​of variable V, respectively. V* represents the normalized variable, which takes values ​​in the range [0,1].

6. The sensitivity analysis method for deformation in arc additive manufacturing according to claim 5, characterized in that, Drawing deformation contour maps of process parameters and analyzing the sensitivity of influencing factors also includes: The slope of the deformation contour lines in the process parameter diagram of arc additive manufacturing is analyzed to evaluate the influence of process parameters on the deformation of arc additive manufacturing on the horizontal and vertical axes. This influence is the sensitivity of arc additive manufacturing deformation to process parameters.

7. The sensitivity analysis method for deformation in arc additive manufacturing according to claim 6, characterized in that, When the slope of the arc additive manufacturing deformation contour line is less than the preset threshold, it indicates that the process parameters corresponding to the vertical axis have a significant impact on the deformation of the arc additive manufacturing component; conversely, when the slope of the arc additive manufacturing deformation contour line is greater than the preset threshold, it indicates that the process parameters corresponding to the horizontal axis have a significant impact on the deformation of the arc additive manufacturing component.

8. A device, characterized in that, Includes memory and processor, wherein: Memory is used to store computer programs that can run on a processor; A processor, configured to, while running the computer program, perform the steps of the sensitivity analysis method for the influence factors of deformation in arc additive manufacturing as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by at least one processor, implements the steps of the sensitivity analysis method for the influence factors of deformation in arc additive manufacturing as described in any one of claims 1-7.

Images