Missing data generation and prediction method for IN718 nickel-based alloy forming based on machine learning

Abstract

The present invention discloses a method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning, including: constructing a core data set for IN718 nickel-based alloy forming and performing data preprocessing; eliminating abnormal data based on data similarity screening criteria and distribution characteristics; constructing a machine learning model for a complete data set of core process parameters; constructing a core process parameter missing data matrix based on an incomplete data set of core process parameters; taking minimizing mean square error as an optimization goal, using gradient optimization to generate the optimal filling value of the missing data of the core process parameters to form a filled data set of the core process parameters; merging the filled data set and the complete data set to construct a machine learning model, evaluating the accuracy of the machine learning model according to the leave-one-out validation method, and outputting the machine learning model. The present invention can predict the core performance indicators of IN718 nickel-based alloy forming and fill in the missing data of the core process parameters, achieve the optimal filling of the missing data, and significantly improve the prediction ability.

Description

Technical Field

[0001] The present invention relates to the field of general artificial intelligence technology and equipment additive manufacturing, and more specifically, to a method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning. Background Art

[0002] IN718 nickel-based alloy is a precipitation hardening nickel-chromium alloy with good high temperature creep strength, fatigue resistance, oxidation resistance and hot corrosion resistance. IN718 nickel-based alloy has excellent mechanical properties and is therefore widely used in key working parts of key equipment in high temperature and high strength environments such as aerospace, petrochemical and nuclear industries.

[0003] Selective laser melting is an additive manufacturing technology based on laser forming that can manufacture complex structures of IN718 nickel-based alloys. At the same time, selective laser melting has attracted widespread attention because it can overcome the shortcomings of low density of parts formed by laser sintering technology and low precision of parts formed by laser cladding technology. However, in selective laser melting technology, the value of core process parameters has a great influence on the overall quality of the formed material, such as porosity, hardness, and surface roughness. For this reason, research on the optimal core process parameters of selective laser melting technology has been favored.

[0004] Furthermore, existing studies have analyzed the values ​​of core process parameters through traditional physical experiments and experimental design methods.

[0005] In the process of implementing the technical method of the embodiment of the present invention, the prior art has the following technical problems:

[0006] Due to the differences in manufacturing equipment in traditional physical experiments and experimental design methods, as well as the high price of equipment, the core process parameters that different researchers focus on are not completely consistent, resulting in incomplete relevant data sets for existing research.

[0007] Considering that the machine learning model can learn the intrinsic relationship between input parameters and output responses, for this reason, machine learning related technologies can be used to fill in the missing data in the incomplete data set, thereby significantly improving the prediction accuracy of the machine learning model and providing a reliable reference model for the optimal selection of core process parameters. Specifically, how to effectively fill in the incomplete data in existing research, and how to integrate the filled data set to construct and evaluate the accuracy of the machine learning model are the research focuses of the present invention. The missing data generation and prediction method based on machine learning designed by the present invention can not only be used for the analysis of the core process parameter values ​​of IN718 nickel-based alloy forming, but also provides a usable data generation structure and prediction model for the optimization of core process parameters of other forming technologies. Summary of the invention

[0008] Aiming at the technical problems that there are missing data in the forming process of IN718 nickel-based alloy in the prior art and the forming mechanism is still unclear, a method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning is studied and designed. The method effectively combines the autonomous learning ability of machine learning with the fast convergence ability of gradient optimization algorithm, determines the optimal filling value of missing data of core process parameters through gradient optimization algorithm, and learns the internal connection of all core process parameter data through machine learning model, providing an accurate reference model for clarifying the forming mechanism of IN718 nickel-based alloy.

[0009] To achieve the above objectives, an embodiment of the present invention provides a method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning, the method comprising:

[0010] Step 1: Based on the historical IN718 nickel-based alloy core data, construct the IN718 nickel-based alloy core data set, and preprocess the core data set based on the similarity of core performance indicators and equipment measurement characteristics;

[0011] Step 2: For the core data set after data preprocessing, remove extremely similar data based on data similarity screening criteria, and remove abnormal data based on data distribution characteristics to form a high-quality core data set;

[0012] Step 3: For high-quality core data sets, determine the complete and incomplete data sets of core process parameters corresponding to each core performance indicator based on data integrity screening, and construct a machine learning model for the complete data set;

[0013] Step 4: For the incomplete data set of core process parameters corresponding to the core performance indicators, construct a missing data matrix of core process parameters, and determine the upper and lower bounds of each missing data in the matrix;

[0014] Step 5: Taking minimizing the mean square error as the optimization goal, gradient optimization is used to generate the optimal filling value of the missing data of the core process parameters to form a filling data set of the core process parameters;

[0015] Step 6: Combine the filled data set and the complete data set to construct a machine learning model, evaluate the accuracy of the machine learning model according to the leave-one-out validation method, output the machine learning model, and provide a reliable and accurate machine learning model for determining the optimal core process parameters for subsequent IN718 nickel-based alloy forming.

[0016] Furthermore, the step 1 includes the following sub-steps:

[0017] Forming IN718 nickel-based alloy by selective laser melting, extracting three core performance indicators, namely porosity, hardness and surface roughness, and four core process parameters corresponding to these three core performance indicators, namely laser power, layer thickness, scanning speed and hatch spacing, from historical IN718 nickel-based alloy core data;

[0018] When the historical IN718 nickel-based alloy core data does not give porosity but only density index, the density index is converted to porosity index value based on the explicit linear relationship between porosity and density index;

[0019] When the measurement units of the three core performance indicators in the existing core literature are inconsistent due to equipment differences, the units of the three core performance indicators in the core data set are unified according to the explicit proportional relationship of the measurement units.

[0020] Furthermore, the step 2 includes the following sub-steps:

[0021] Assuming that the four core process parameter values ​​in each set of data constitute a 4-dimensional vector, in the data similarity screening criteria, if the four core process parameters of two sets of data are slightly different, that is, the Euclidean distance between the 4-dimensional vectors corresponding to the two sets of data is less than the set minimum Euclidean distance threshold, they are considered to be extremely similar data, and any one of the two sets of data is deleted from the core data set; wherein, the calculation formula for the set minimum Euclidean distance threshold is:

[0022] ,

[0023] In the above formula, The Euclidean distance threshold to determine whether all core process parameters in two sets of data are extremely similar; It is a vector composed of the maximum values ​​of the four core process parameters in the core data set after data preprocessing; It is a vector composed of the minimum values ​​of the four core process parameters in the core data set after data preprocessing;

[0024] Calculate the mean and standard deviation of each core performance indicator in the core data set after deleting any one of the two data sets, and remove the abnormal data in the core data set after deleting any one of the two data sets according to the Laida criterion, and reduce the original core data set to a high-quality core data set; among them, the Laida criterion will be located in Data other than the above are considered abnormal data. and are the mean and standard deviation of the data respectively.

[0025] Furthermore, the step three includes the following sub-steps:

[0026] If there are valid values ​​for the four core process parameters of a set of data for the current core performance indicator, the set of data is considered complete and is stored in the core process parameter complete data set;

[0027] If there are invalid values ​​for the four core process parameters of a set of data for the current core performance indicator, the set of data is considered incomplete and is stored in the incomplete core process parameter data set;

[0028] The core process parameter values ​​in the complete data set are taken as input parameters, and the three core performance indicators are taken as three output response values. Three machine learning models of the relationship between the reaction input parameters and the output response are constructed respectively. The specific expressions are as follows:

[0029] ,

[0030] In the above formula, For the The machine learning model prediction values ​​of core performance indicators, among which, When it is equal to 1, 2, and 3, it corresponds to porosity, hardness, and surface roughness, respectively; For the The weight parameters of the core process parameters; Indicates Group Data The basis functions of express and The Euclidean distance between .

[0031] Furthermore, the step 4 includes the following sub-steps:

[0032] In the core process parameter missing data matrix, the valid values ​​of the core process parameters in the incomplete data set are fixed, and the core process parameters with invalid values ​​in each group of data are used as input variables in turn according to the data set sequence and the arrangement sequence of the core process parameters;

[0033] Calculate the maximum and minimum values ​​of each core process parameter corresponding to all data in the high-quality core data set, and use the maximum value as the upper limit of the core process parameter value, and the minimum value as the lower limit of the core process parameter value.

[0034] Furthermore, in step 5, minimizing the mean square error is taken as the optimization goal, and the calculation formula of the mean square error is as follows:

[0035] ,

[0036] In the above formula, For the missing data matrix of core process parameters No. The mean square error of the prediction of the core performance indicators, For the missing data matrix of core process parameters No. The first The machine learning prediction values ​​of the core performance indicators, For the missing data matrix of core process parameters No. The first The true response value of the core performance indicators; The missing data matrix of the core process parameters Group data.

[0037] Furthermore, in step 5, the optimal filling value of the missing data of the core process parameters is generated by using gradient optimization to form a filling data set of the core process parameters, including the following sub-steps:

[0038] In the gradient optimization framework, the mean square error of the missing data matrix of the core process parameters of the machine learning model is taken as the optimization target, and the missing data matrix is ​​taken as the input matrix parameter. The interior point method is constructed to solve this single-objective parameter optimization problem.

[0039] The optimal filling values ​​obtained by the interior point method are input into the missing data matrix of the core process parameters in sequence to form a filling data set of the core process parameters.

[0040] Furthermore, the step six includes the following sub-steps:

[0041] Taking the combined dataset of the padded dataset and the high-quality complete dataset as the modeling sample set, updating the machine learning model of each core performance indicator according to the machine learning model expression;

[0042] In the leave-one-out validation method, each sample data in the modeling sample set is traversed, and the absolute value of the difference between the corresponding machine learning model prediction value and the true response value of the core performance indicator is calculated respectively. The accuracy of the updated machine learning model evaluated by the leave-one-out validation method is further calculated, and an explicit expression of the machine learning model that can predict the key core performance indicators for any combination of core process parameter values ​​is given.

[0043] In a second aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning are implemented.

[0044] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the IN718 nickel-based alloy forming missing data generation and prediction method based on machine learning in the aforementioned embodiment are implemented.

[0045] In summary, compared with the prior art, the above technical solution conceived by the present invention provides a method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning, which mainly has the following beneficial effects:

[0046] 1. The data similarity screening criteria are used to effectively eliminate extremely similar data, avoiding the phenomenon of morbid correlation coefficient matrix caused by extremely similar data when building machine learning models, and improving the construction stability of machine learning models.

[0047] 2. Based on the Laida rule, abnormal data is effectively eliminated, avoiding the possibility of a significant decrease in the prediction performance of the machine learning model due to a small number of abnormal data, and improving the prediction accuracy of the machine learning model.

[0048] 3. The missing data of core process parameters are optimally filled through the interior point method with extremely fast convergence speed, which improves the efficiency of missing data generation.

[0049] 4. The merged data set effectively increases the data volume, making the updated machine learning model more accurate. At the same time, the leave-one-out validation method ensures the reproducibility of the prediction accuracy of the machine learning model, making the updated machine learning model a reference model for the study of the forming mechanism of IN718 nickel-based alloy.

[0050] 5. The present invention can efficiently generate missing data of the core process parameters of IN718 nickel-based alloy forming and accurately predict key core performance indicators, which accelerates the theoretical research progress on IN718 nickel-based alloy forming technology and can be widely used to fill in missing data of core process parameters and predict key core performance indicators of various forming technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flowchart of a method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning provided by the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0053] See also Figure 1 The present invention provides a method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning, which is applicable to the problem of IN718 nickel-based alloy forming. Specifically, the method mainly includes the following steps:

[0054] Step 1: Based on the historical IN718 nickel-based alloy core data, the IN718 nickel-based alloy core data set is constructed, and the core data set is preprocessed based on the similarity of core performance indicators and equipment measurement characteristics. The specific steps include:

[0055] (11) Forming IN718 nickel-based alloy by selective laser melting, extracting three core performance indicators, namely porosity, hardness and surface roughness, and four core process parameters corresponding to the three core performance indicators, namely laser power, layer thickness, scanning speed and hatch spacing, from the historical IN718 nickel-based alloy core data;

[0056] (12) When porosity is not given in the historical IN718 nickel-based alloy core data but only density index is given, the density index is converted into porosity index value based on the explicit linear relationship between porosity and density index;

[0057] (13) When the measurement units of the three core performance indicators in the historical IN718 nickel-based alloy core data are inconsistent due to equipment differences, the units of the three core performance indicators in the core data set are unified according to the explicit proportional relationship of the measurement units;

[0058] Step 2: For the core data set after data preprocessing, remove extremely similar data based on the data similarity screening criteria, and remove abnormal data based on the data distribution characteristics to form a high-quality core data set. The specific steps are as follows:

[0059] (21) Assuming that the four core process parameter values ​​in each set of data form a 4-dimensional vector, in the data similarity screening criteria, if the four core process parameters of two sets of data are slightly different, that is, the Euclidean distance between the 4-dimensional vectors corresponding to the two sets of data is less than the set minimum Euclidean distance threshold, they are considered to be extremely similar data, and any one of the two sets of data is deleted from the core data set. The calculation formula for the set minimum Euclidean distance threshold is:

[0060] ;

[0061] In the above formula, The Euclidean distance threshold to determine whether all core process parameters in two sets of data are extremely similar; It is a vector composed of the maximum values ​​of the four core process parameters in the core data set after data preprocessing; It is a vector composed of the minimum values ​​of the four core process parameters in the core data set after data preprocessing;

[0062] (22) Calculate the mean and standard deviation of each core performance indicator in the core data set after deleting any one of the two data sets, and remove the abnormal data in the core data set after deleting any one of the two data sets according to the Laida criterion, so as to reduce the original core data set to a high-quality core data set; among which, the Laida criterion will be located in Data other than the above are considered abnormal data. and are the mean and standard deviation of the data respectively.

[0063] Step 3: For high-quality core data sets, determine the complete and incomplete data sets of core process parameters corresponding to each core performance indicator based on data integrity screening, and construct a machine learning model for the complete data set. The specific steps are as follows:

[0064] (31) If there are valid values ​​for the four core process parameters of a set of data for the current core performance indicator, the set of data is considered complete and is stored in the core process parameter complete data set;

[0065] (32) If there are invalid values ​​for the four core process parameters of a set of data for the current core performance indicator, the set of data is considered incomplete and is stored in the incomplete core process parameter data set;

[0066] (33) The core process parameter values ​​in the complete data set are taken as input parameters, and the three core performance indicators are taken as three output response values. Three machine learning models of the relationship between the reaction input parameters and the output response are constructed respectively. The specific expressions are as follows:

[0067] ;

[0068] In the above formula, For the The machine learning model prediction values ​​of core performance indicators, among which, When it is equal to 1, 2, and 3, it corresponds to porosity, hardness, and surface roughness, respectively; For the The weight parameters of the core process parameters; Indicates Group Data The basis functions of express and The Euclidean distance between .

[0069] Step 4: For the incomplete data set of core process parameters corresponding to the core performance indicators, construct a missing data matrix of core process parameters and determine the upper and lower bounds of each missing data in the matrix. The specific steps are as follows:

[0070] (41) In the core process parameter missing data matrix, the valid values ​​of the core process parameters in the incomplete data set are fixed, and the core process parameters with invalid values ​​in each set of data are used as input variables in turn according to the order of the data sets and the arrangement order of the core process parameters;

[0071] (42) Calculate the maximum and minimum values ​​of each core process parameter corresponding to all data in the high-quality core data set, and use the maximum value as the upper limit of the core process parameter value, and the minimum value as the lower limit of the core process parameter value.

[0072] Step 5: The calculation formula of the mean square error corresponding to the optimization goal of minimizing the mean square error is as follows:

[0073] ;

[0074] In the above formula, For the missing data matrix of core process parameters No. The mean square error of the prediction of the core performance indicators, For the missing data matrix of core process parameters No. The first The machine learning prediction values ​​of the core performance indicators, For the missing data matrix of core process parameters No. The first The true response value of the core performance indicators; The missing data matrix of the core process parameters Group data.

[0075] Furthermore, gradient optimization is used to generate the optimal filling value of the missing data of the core process parameters to form a filling data set of the core process parameters. The specific steps are as follows:

[0076] (51) In the gradient optimization framework, the mean square error of the missing data matrix of the core process parameters of the machine learning model is taken as the optimization target, and the missing data matrix is ​​taken as the input matrix parameter. The interior point method is constructed to solve this single-objective parameter optimization problem.

[0077] (52) The optimal filling values ​​obtained by the interior point method are sequentially input into the missing data matrix of the core process parameters to form a filling data set of the core process parameters.

[0078] Step 6: Combine the filled data set and the complete data set to construct a machine learning model, evaluate the accuracy of the machine learning model according to the leave-one-out validation method, output the machine learning model, and provide a reliable machine learning model for determining the optimal core process parameters for subsequent IN718 nickel-based alloy forming. The specific steps are as follows:

[0079] (61) Using the combined dataset of the padded dataset and the high-quality complete dataset as the modeling sample set, updating the machine learning model of each core performance indicator according to the machine learning model expression in step 3;

[0080] (62) In the leave-one-out validation method, each sample data in the modeling sample set is traversed, and the absolute value of the difference between the corresponding machine learning model prediction value and the true response value of the core performance indicator is calculated respectively. The accuracy of the updated machine learning model evaluated by the leave-one-out validation method is further calculated, and an explicit expression of the machine learning model that can predict the key core performance indicators for any combination of core process parameter values ​​is given.

[0081] Example

[0082] See also Figure 1 This embodiment is explained with the key core process parameter data provided by the core data of IN718 nickel-based alloy in the past five years, and these data are as follows:

[0083] Core performance indicators Porosity hardness Surface roughness Complete data volume 220 108 114 Incomplete data volume 72 43 8

[0084] After the above data are processed by the IN718 nickel-based alloy forming missing data generation and prediction method based on machine learning in the embodiment of the present invention, a machine learning model is constructed. In order to further illustrate this embodiment, the machine learning model constructed by the initial complete core data set and the merged data set in this embodiment is compared to determine the coefficient As the accuracy evaluation index, the experimental results are shown in the table below. The core process parameter missing data generation method of this embodiment effectively improves the prediction accuracy of the machine learning model, which reflects the effectiveness of the data generation method of this embodiment and the accuracy of the machine learning model. It can be considered that the method of this embodiment has a good reference role in the theoretical research of IN718 nickel-based alloy forming.

[0085] Machine Learning Model Accuracy Comparison Table

[0086] algorithm Initial complete core dataset Merge Datasets Porosity 0.73 0.99 hardness 0.15 0.987 Surface roughness 0.80 0.92

[0087] The present invention provides a method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning. The method effectively improves the prediction ability of the machine learning model through a data generation method based on gradient optimization, and provides an effective reference model for the forming mechanism of IN718 nickel-based alloy.

[0088] Another embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the IN718 nickel-based alloy forming missing data generation and prediction method based on machine learning in the aforementioned embodiment are implemented.

[0089] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning in the aforementioned embodiment are implemented.

[0090] It will be easily understood by those skilled in the art that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for generating and predicting missing data for IN718 nickel-based alloy forming based on machine learning, characterized in that: The method comprises: Step 1: Based on the historical IN718 nickel-based alloy core data, construct the IN718 nickel-based alloy core data set, and preprocess the core data set based on the similarity of core performance indicators and equipment measurement characteristics; Step 2: For the core data set after data preprocessing, remove extremely similar data based on data similarity screening criteria, and remove abnormal data based on data distribution characteristics to form a high-quality core data set; Step 3: For high-quality core data sets, determine the complete and incomplete data sets of core process parameters corresponding to each core performance indicator based on data integrity screening, and construct a machine learning model for the complete data set; Step 4: For the incomplete data set of core process parameters corresponding to the core performance indicators, construct a missing data matrix of core process parameters, and determine the upper and lower bounds of each missing data in the matrix; Step 5: Taking minimizing the mean square error as the optimization goal, gradient optimization is used to generate the optimal filling value of the missing data of the core process parameters to form a filling data set of the core process parameters; Step 6: Combine the filled data set and the complete data set to construct a machine learning model, evaluate the accuracy of the machine learning model according to the leave-one-out validation method, output the machine learning model, and provide a reliable and accurate machine learning model for determining the optimal core process parameters for subsequent IN718 nickel-based alloy forming; In step 5, the optimization goal is to minimize the mean square error, and the calculation formula of the mean square error is as follows: ， In the above formula, For the missing data matrix of core process parameters No. The mean square error of the prediction of the core performance indicators, For the missing data matrix of core process parameters No. The first The machine learning prediction values ​​of the core performance indicators, For the missing data matrix of core process parameters No. The first The true response value of the core performance indicators; The missing data matrix of the core process parameters Group data; In step 5, gradient optimization is used to generate the optimal filling value of the missing data of the core process parameters to form a filling data set of the core process parameters, including the following sub-steps: In the gradient optimization framework, the mean square error of the missing data matrix of the core process parameters of the machine learning model is taken as the optimization target, and the missing data matrix is ​​taken as the input matrix parameter. The interior point method is constructed to solve this single-objective parameter optimization problem. The optimal filling values ​​obtained by the interior point method are input into the missing data matrix of the core process parameters in sequence to form a filling data set of the core process parameters.

2. The method according to claim 1, characterized in that The step 1 includes the following sub-steps: Forming IN718 nickel-based alloy by selective laser melting, extracting three core performance indicators, namely porosity, hardness and surface roughness, and four core process parameters corresponding to these three core performance indicators, namely laser power, layer thickness, scanning speed and hatch spacing, from historical IN718 nickel-based alloy core data; When the historical IN718 nickel-based alloy core data does not give porosity but only density index, the density index is converted to porosity index value based on the explicit linear relationship between porosity and density index; When the measurement units of the three core performance indicators in the existing core literature are inconsistent due to equipment differences, the units of the three core performance indicators in the core data set are unified according to the explicit proportional relationship of the measurement units.

3. The method according to claim 1, characterized in that The step 2 includes the following sub-steps: Assuming that the four core process parameter values ​​in each set of data constitute a 4-dimensional vector, in the data similarity screening criteria, if the four core process parameters of two sets of data are slightly different, that is, the Euclidean distance between the 4-dimensional vectors corresponding to the two sets of data is less than the set minimum Euclidean distance threshold, they are considered to be extremely similar data, and any one of the two sets of data is deleted from the core data set; wherein, the calculation formula for the set minimum Euclidean distance threshold is: ， In the above formula, The Euclidean distance threshold is used to determine whether all core process parameters in two sets of data are extremely similar; It is a vector composed of the maximum values ​​of the four core process parameters in the core data set after data preprocessing; It is a vector composed of the minimum values ​​of the four core process parameters in the core data set after data preprocessing; Calculate the mean and standard deviation of each core performance indicator in the core data set after deleting any one of the two data sets, and remove the abnormal data in the core data set after deleting any one of the two data sets according to the Laida criterion, and reduce the original core data set to a high-quality core data set; among them, the Laida criterion will be located in Data other than the above are considered abnormal data. and are the mean and standard deviation of the data respectively.

4. The method according to claim 1, characterized in that The step three includes the following sub-steps: If there are valid values ​​for the four core process parameters of a set of data for the current core performance indicator, the set of data is considered complete and is stored in the core process parameter complete data set; If there are invalid values ​​for the four core process parameters of a set of data for the current core performance indicator, the set of data is considered incomplete and is stored in the incomplete core process parameter data set; The core process parameter values ​​in the complete data set are taken as input parameters, and the three core performance indicators are taken as three output response values. Three machine learning models of the relationship between the reaction input parameters and the output response are constructed respectively. The specific expressions are as follows: ， In the above formula, For the The machine learning model prediction values ​​of core performance indicators, among which, When it is equal to 1, 2, and 3, it corresponds to porosity, hardness, and surface roughness, respectively; For the The weight parameters of the core process parameters; Indicates Group Data The basis functions of express and The Euclidean distance between .

5. The method according to claim 1, wherein the step 4 comprises the following sub-steps: In the core process parameter missing data matrix, the valid values ​​of the core process parameters in the incomplete data set are fixed, and the core process parameters with invalid values ​​in each group of data are used as input variables in turn according to the data set sequence and the arrangement sequence of the core process parameters; Calculate the maximum and minimum values ​​of each core process parameter corresponding to all data in the high-quality core data set, and use the maximum value as the upper limit of the core process parameter value, and the minimum value as the lower limit of the core process parameter value.

6. The method according to claim 4, characterized in that The step six includes the following sub-steps: Taking the combined dataset of the padded dataset and the high-quality complete dataset as the modeling sample set, updating the machine learning model of each core performance indicator according to the machine learning model expression; In the leave-one-out validation method, each sample data in the modeling sample set is traversed, and the absolute value of the difference between the corresponding machine learning model prediction value and the true response value of the core performance indicator is calculated respectively. The accuracy of the updated machine learning model evaluated by the leave-one-out validation method is further calculated, and an explicit expression of the machine learning model that can predict the key core performance indicators for any combination of core process parameter values ​​is given.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images