Method for deformation control of binder jet manufactured metal parts

Abstract

The present application relates to a kind of binder jetting manufacturing metal parts deformation control method.The deformation control method is by comprehensively collecting the deformation data of different process manufacturing metal parts, establish the underlying database supporting deformation prediction, ensure the high confidence of prediction result;Through data analysis, screen the key factor influencing deformation, effectively reduce the calculation amount of deformation prediction, ensure the efficient operation of deformation control method;Through model creation, establish the quantitative coupling relationship between each key factor and the deformation of forming part, ensure that deformation prediction is simple and easy to operate.Compared with the prior art, the above-mentioned binder jetting manufacturing metal parts deformation control method comprehensively considers the material properties, process parameters and other factors influencing deformation, and the calculation process is simple, the result confidence is high, and the engineering guidance is great.

Description

Technical Field

[0001] This invention relates to the field of binder spraying additive manufacturing technology, and in particular to a method for controlling the deformation of metal parts manufactured by binder spraying additive manufacturing. Background Technology

[0002] Binder spraying technology is a method based on the principle of micro-spraying, which uses liquid binder to bond and solidify evenly spread powder layer by layer to form parts. Initially used mainly for casting mold making, this technology has been widely adopted in the field of metal forming due to its advantages such as high efficiency, low cost, wide range of formable material systems, good surface quality, and no need for support structures. It has gradually expanded from iron-based materials to titanium alloys, high-temperature alloys, tungsten alloys, and even active metal materials such as aluminum and magnesium.

[0003] The process of binder spraying to manufacture metal parts typically includes powder bed preparation, preform printing, preform curing, preform debinding, and high-temperature sintering. Because the microstructure of the metal part preform changes from a loosely packed powder state to a tightly bonded state after sintering, it experiences dimensional shrinkage of over 20%. This often results in problems such as large deformation and poor precision in binder spraying of metal parts, limiting the engineering application of this technology. Summary of the Invention

[0004] Therefore, it is necessary to provide a method for deformation control of metal parts manufactured by adhesive spraying that has a wide range of applicable materials and structures, is simple and effective, and yields accurate results.

[0005] A method for controlling deformation during the manufacture of metal parts by adhesive spraying includes the following steps:

[0006] Based on experimental test data, a process database is obtained; the process database includes raw material characteristics, metal part dimensions, binder spraying process parameters, porosity characteristics of the metal part blank manufactured by binder spraying, and deformation amount of the metal part manufactured by binder spraying.

[0007] Correlation analysis was performed on the different raw material characteristics, binder spraying process parameters and porosity characteristics of metal part blanks manufactured by binder spraying in the process database to screen out key factors affecting deformation as the first input feature;

[0008] Correlation analysis was performed on the different metal component sizes, different porosity characteristics of the metal component blanks manufactured by binder spraying, and the deformation of the metal components manufactured by binder spraying in the process database to screen out the key factors affecting deformation as the second input features.

[0009] Based on the liquid phase sintering theory, a preliminary deformation prediction model is created using the first input feature and the second input feature as input quantities and a machine learning algorithm.

[0010] Using the Nash efficiency coefficient as a criterion, the preliminary deformation prediction model is iteratively optimized to obtain an optimized deformation prediction model.

[0011] In one embodiment, after obtaining the optimized deformation prediction model, the step further includes:

[0012] Based on the dimensional data of the component to be manufactured, the deformation data of the component to be manufactured is obtained using the optimized deformation prediction model;

[0013] Based on the deformation data of the part to be manufactured, a deformation inverse compensation design is performed on the printing model of the part to be manufactured to obtain a precision printing model.

[0014] In one embodiment, the raw material characteristics include the particle size of the metal powder, the flowability of the metal powder, the morphology of the metal powder, the composition of the binder, the viscosity of the binder, and the surface tension of the binder.

[0015] In one embodiment, the adhesive spraying process parameters include printing layer thickness, powder spreading speed, adhesive saturation, and adhesive spraying amount per layer.

[0016] In one embodiment, the porosity characteristics of the metal part blank manufactured by the binder spraying include porosity, pore space distribution, and pore size.

[0017] In one embodiment, the deformation of the metal part manufactured by the adhesive spraying includes deformation data in the horizontal direction and deformation data in the vertical direction.

[0018] In one embodiment, in the process database, each parameter of the raw material characteristics has at least 5 data points that change in a stepwise manner, each parameter of the binder spraying process parameters has at least 5 data points that change in a stepwise manner, each parameter of the porosity characteristics of the metal part blank manufactured by binder spraying has at least 50 measured data points, and each parameter of the deformation amount of the metal part blank manufactured by binder spraying has at least 50 different data points.

[0019] In one embodiment, the step of performing a correlation analysis on the characteristics of different raw materials, binder spraying process parameters, and porosity characteristics of the metal part blank manufactured by binder spraying to obtain key parameters as the first input feature includes the following steps:

[0020] The raw material characteristics, binder spraying process parameters, and porosity characteristics of the metal part blank manufactured by binder spraying in the process database are each considered as the first set of independent and dependent variables.

[0021] The first set of independent-dependent variables is divided into a first training set and a first validation set;

[0022] Based on the data from the first training set, a model of the effect of independent variables on dependent variables was created and fitted using the stepwise regression method.

[0023] In the aforementioned action model, by setting independent variables that have the least significant effect on the dependent variable at each step, key factors affecting the deformation are obtained as the first input features.

[0024] Using the independent variables in the first validation set as inputs, the dependent variable in the first validation set is predicted using the action model to obtain the prediction result.

[0025] The mean squared error is calculated by comparing the measured results and the predicted results of the dependent variable in the first validation set.

[0026] The interaction model is cross-validated using the mean squared error as the criterion.

[0027] In one embodiment, the step of performing a correlation analysis on different metal part sizes, porosity characteristics of the metal part blank manufactured by binder spraying, and deformation of the metal part blank manufactured by binder spraying to screen out key parameters as second input features includes:

[0028] The dimensions of the metal parts, the porosity characteristics of the metal part blanks manufactured by binder spraying, and the deformation of the metal parts manufactured by binder spraying in the process database are each considered as the second set of independent and dependent variables.

[0029] The second group of independent-dependent variables is divided into a second training set and a second validation set;

[0030] Based on the data from the second training set, a model of the effect of independent variables on dependent variables was created and fitted using the stepwise regression method.

[0031] In the aforementioned action model, by setting independent variables that have the least significant effect on the dependent variable at each step, key factors influencing deformation are obtained as second input features.

[0032] Using the independent variables in the second validation set as inputs, the dependent variable in the second validation set is predicted using the action model to obtain the prediction result;

[0033] The mean squared error is calculated by comparing the measured and predicted results of the dependent variable in the second validation set.

[0034] The interaction model is cross-validated using the mean squared error as the criterion.

[0035] In one embodiment, the step of creating a preliminary deformation prediction model using a machine learning algorithm includes the following steps:

[0036] Using the first input feature and the second input feature as the input layer, the liquid phase sintering theory as the hidden layer, and the deformation amount of the metal parts manufactured by binder spraying as the output layer, a neural network is constructed.

[0037] Determine the weights and biases of the input layer neurons;

[0038] Based on the input layer data, the output layer data is obtained by performing forward propagation calculations on the neural network;

[0039] The loss function is obtained by comparing the output layer data with the measured values ​​of the deformation of the metal parts manufactured by adhesive spraying;

[0040] The gradient of each layer is obtained by backpropagation calculation using the neural network.

[0041] Optimize the neural network parameters and minimize the error using the gradient descent method;

[0042] Return to the steps that yield the output layer data, and continue until the neural network converges to obtain a preliminary deformation prediction model.

[0043] The aforementioned variable control method for manufacturing metal parts using binder spraying comprehensively collects deformation data from metal parts manufactured using different processes, establishing a foundational database to support deformation prediction and ensuring high confidence in the prediction results. Secondly, through data analysis, key factors affecting deformation are screened, effectively reducing the computational load for deformation prediction and ensuring efficient operation of the deformation control method. Subsequently, through model creation, a quantitative coupling relationship between each key factor and the deformation amount of the formed part is established, ensuring ease of deformation prediction operation. Compared to existing technologies, the aforementioned deformation control method for manufacturing metal parts using binder spraying comprehensively considers factors such as material properties and process parameters affecting deformation, and its calculation process is simple, with high confidence in the results, offering significant engineering guidance value. Attached Figure Description

[0044] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0045] Figure 1 This is a schematic flowchart of a method for controlling the deformation of metal parts manufactured by adhesive spraying in a preferred embodiment of the present invention.

[0046] Figure 2 for Figure 1 The flowchart shown below illustrates the deformation control method for manufacturing metal parts using adhesive spraying, following step S500.

[0047] Figure 3 for Figure 1 The flowchart of step S200 in the deformation control method for manufacturing metal parts by adhesive spraying is shown.

[0048] Figure 4 for Figure 1 The flowchart of step S300 in the deformation control method for manufacturing metal parts by adhesive spraying is shown.

[0049] Figure 5 for Figure 1 The flowchart of step S400 in the method for controlling the deformation of metal parts manufactured by adhesive spraying is shown. Detailed Implementation

[0050] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0052] When describing positional relationships, unless otherwise specified, when an element is referred to as being "on" another element, it may be directly on the other element or there may be intermediate elements. It is also understood that when an element is referred to as being "between" two elements, it may be the only one between the two elements, or there may be one or more intermediate elements.

[0053] When using the terms “including,” “having,” and “comprising” as described herein, another component may be added unless explicitly qualifying terms such as “only,” “consisting of,” etc. are used. Unless otherwise stated, singular terms may include plural forms and should not be construed as having a quantity of one.

[0054] Please see Figure 1 The deformation control method for manufacturing metal parts by adhesive spraying in the preferred embodiment of the present invention is applicable to various metal material systems that can be formed by adhesive spraying, such as titanium alloys, high-temperature alloys, and tungsten alloys, and includes steps S100 to S500.

[0055] Step S100: Based on experimental test data, obtain the process database. The process database includes raw material properties, metal part dimensions, binder spraying process parameters, porosity characteristics of the metal part blank manufactured by binder spraying, and deformation amount of the metal part manufactured by binder spraying.

[0056] Specifically, the characteristics of the raw materials include the particle size of the metal powder, the flowability of the metal powder, the morphology of the metal powder, the composition of the binder, the viscosity of the binder, and the surface tension of the binder.

[0057] Specifically, the adhesive spraying process parameters include printing layer thickness, powder spreading speed, adhesive saturation, and adhesive spraying amount per layer.

[0058] Specifically, the porosity characteristics of metal component blanks manufactured by binder spraying include porosity, pore space distribution, and pore size.

[0059] Specifically, the deformation of metal parts manufactured by adhesive spraying includes deformation data in the horizontal direction and deformation data in the vertical direction.

[0060] Specifically, using the morphology of the raw material metal powder, the average particle size distribution range, the printing thickness of the binder spray, the powder spreading speed, the binder saturation, and the powder spreading speed as factors, an orthogonal experiment is designed to manufacture metal part blanks. The internal porosity of the prepared blanks is characterized by μCT, and the dimensional changes after sintering are tested using a coordinate measuring machine to obtain experimental test data, thereby creating a process database.

[0061] Step S200: Correlation analysis is performed on the different raw material characteristics, binder spraying process parameters and porosity characteristics of the metal part blank manufactured by binder spraying in the process database to screen out the key factors affecting deformation as the first input features.

[0062] Among them, the different raw material characteristics and binder spraying process parameters in the process database are used as one set of data, and the porosity characteristics of the metal part blank manufactured by binder spraying in the process database are used as another set of data. Correlation analysis is performed on the two sets of data.

[0063] Step S300: Correlation analysis is performed on the different metal part sizes, different porosity characteristics of the metal part blanks manufactured by binder spraying and the deformation amount of the metal parts manufactured by binder spraying in the process database, so as to screen out the key factors affecting deformation as the second input features.

[0064] Among them, the different metal parts dimensions and different porosity characteristics of the metal parts blanks manufactured by binder spraying in the process database are used as one set of data, and the deformation of the metal parts manufactured by binder spraying in the process database is used as another set of data. Correlation analysis is performed on the two sets of data.

[0065] Step S400: Based on the liquid phase sintering theory, a preliminary deformation prediction model is created using the first input feature and the second input feature as input quantities and a machine learning algorithm.

[0066] Step S500: Using the Nash efficiency coefficient as a criterion, iteratively optimize the preliminary deformation prediction model to obtain an optimized deformation prediction model.

[0067] By executing steps S100 to S500, an optimized deformation prediction model can be obtained. The prediction results of the optimized deformation prediction model have a high confidence level. At the same time, the deformation data output by the optimized deformation prediction model after manufacturing can effectively guide the subsequent processing of binder spraying additive manufacturing metal parts, thereby achieving precision manufacturing.

[0068] Step S100 is executed to comprehensively collect deformation data of metal parts manufactured using different processes, establishing a foundational database to support deformation prediction and ensuring high confidence in the prediction results. Steps 200 and S300 are executed to use data analysis to screen out key factors affecting deformation, effectively reducing the computational load of deformation prediction and ensuring efficient operation of the deformation control method. Steps S400 and S500 are executed to create and optimize the model, obtaining an optimized deformation prediction model and establishing a quantitative coupling relationship between each key factor and the deformation amount of metal parts manufactured by binder spraying, ensuring ease of deformation prediction operation.

[0069] Practical testing has proven that the deformation control method for manufacturing metal parts using the aforementioned binder spraying technique can effectively predict the deformation trends and amounts of various materials and structures, with the deviation between the predicted results and actual measurements being less than 10%. Therefore, compared to existing technologies, this method for controlling the deformation of metal parts using binder spraying comprehensively considers factors such as material properties and process parameters that affect deformation. Furthermore, the calculation process is simple, and the results have high confidence levels. It can effectively guide the precision forming of metal parts during binder spraying additive manufacturing, and its engineering guidance significance is immense.

[0070] In some embodiments, the process database contains at least five data points showing a stepped variation for each parameter of the raw material properties, at least five data points showing a stepped variation for each parameter of the binder spraying process parameters, at least 50 measured data points for each parameter of the porosity characteristics of the binder-sprayed metal part blank, and at least 50 different data points for each parameter of the deformation amount of the binder-sprayed metal part blank. This ensures that the process database has sufficient data to complete the creation of the subsequent preliminary deformation prediction model, thereby ensuring that high-confidence results can be obtained by optimizing the deformation prediction model.

[0071] Please refer to the following: Figure 2In some embodiments, after step S500, steps S601 and S602 are further included:

[0072] Step S601: Based on the dimensional data of the part to be manufactured, obtain the deformation data of the part to be manufactured using an optimized deformation prediction model.

[0073] Step S602: Perform deformation inverse compensation design on the printing model of the part to be manufactured based on the deformation data of the part to be manufactured, so as to obtain a precision printing model.

[0074] By executing steps S601 and S602, deformation and reverse compensation design can be performed on the printed model of the part to be manufactured to obtain a precision printed model, which can then be used to guide the forming of metal parts during the binder jet additive manufacturing process.

[0075] Furthermore, practical tests have proven that the metal part blanks obtained by binder jet additive manufacturing using the above-mentioned precision printed model can achieve a dimensional accuracy of ±0.3mm after debinding and sintering, thus realizing high-precision manufacturing.

[0076] Please refer to the following: Figure 3 In some embodiments, step S200 includes steps S210 to S270.

[0077] Step S210: The raw material characteristics, binder spraying process parameters, and porosity characteristics of the metal part blank manufactured by binder spraying in the process database are each other as the first set of independent-dependent variables.

[0078] That is, in the first set of independent-dependent variables, the independent variables are the raw material characteristics and binder spraying process parameters in the process database, and the dependent variable is the porosity characteristics of the metal part blank manufactured by binder spraying in the process database.

[0079] Step S220: Divide the first set of independent-dependent variables into a first training set and a first validation set.

[0080] Step S230: Based on the data from the first training set, a model of the effect of the independent variable on the dependent variable is created and fitted using the stepwise regression method.

[0081] The model is created by fitting different independent and dependent variables using stepwise regression.

[0082] Step S240: In the action model, by setting the independent variables that have the least significant effect on the dependent variable at each step, the key factors affecting the deformation are obtained as the first input features. That is, in the action model, the independent variables that need to be eliminated at each step are set, and the independent variables that need to be eliminated are precisely those independent variables that have the least significant effect on the dependent variable.

[0083] Step S250: Using the independent variables in the first validation set as inputs, the dependent variable in the first validation set is predicted using the action model to obtain the prediction result.

[0084] Step S260: Compare the measured results of the dependent variable in the first validation set with the predicted results obtained in step S250 to calculate the mean square error.

[0085] Step S270: Cross-validate the model using the mean squared error as the criterion.

[0086] Thus, by executing steps S210 to S270 above, using stepwise regression analysis, based on the raw material characteristics, binder spraying process parameters, and porosity characteristics of the metal part blank manufactured by binder spraying in the process database, the key factors affecting deformation are obtained as the first input features, so as to provide a more accurate input quantity for the subsequent creation of the preliminary deformation prediction model. The execution of steps S250 to S270 can improve the confidence of the first input features.

[0087] Of course, in other embodiments, linear regression, logistic regression, or other regression analysis methods can also be used to obtain the key factors affecting deformation.

[0088] Please refer to the following: Figure 4 In some embodiments, step S300 includes steps S310 to S370.

[0089] Step S310: The dimensions of the metal parts in the process database, the porosity characteristics of the metal part blank manufactured by binder spraying, and the deformation of the metal parts manufactured by binder spraying are each other as the second set of independent-dependent variables.

[0090] That is, in the second set of independent-dependent variables, the independent variables are the dimensions of the metal parts and the porosity characteristics of the metal part blanks manufactured by binder spraying in the process database, and the dependent variable is the deformation of the metal parts manufactured by binder spraying in the process database.

[0091] Step S320: Divide the second group of independent-dependent variables into a second training set and a second validation set.

[0092] Step S330: Based on the data from the second training set, a model of the effect of independent variables on dependent variables is created and fitted using the stepwise regression method.

[0093] The model is created by fitting different independent and dependent variables using stepwise regression.

[0094] Step S340: In the action model, by setting the independent variables that have the least significant effect on the dependent variable at each step, the key factors affecting the deformation are obtained as the second input features. That is, in the action model, the independent variables that need to be eliminated at each step are set, and the independent variables that need to be eliminated are those that have the least significant effect on the dependent variable.

[0095] Step S350: Using the independent variables in the second validation set as inputs, the dependent variable in the second validation set is predicted using the action model to obtain the prediction result.

[0096] Step S360: Compare the measured results and predicted results of the dependent variable in the first validation set to calculate the mean square error.

[0097] Step S370. Perform cross-validation on the action model using the mean squared error as the criterion.

[0098] Thus, by executing steps S310 to S370 above, using stepwise regression analysis, based on the dimensions of the metal parts in the process database, the porosity characteristics of the metal part blank manufactured by binder spraying, and the deformation of the metal parts manufactured by binder spraying, the key factors affecting deformation are obtained as the second input features, so as to provide a more accurate input for the creation of the subsequent preliminary deformation prediction model. The execution of steps S350 to S370 can improve the confidence of the second input features.

[0099] Of course, in other embodiments, linear regression, logistic regression, or other regression analysis methods can also be used to obtain the key factors affecting deformation.

[0100] Please refer to the following: Figure 5 In some embodiments, step S400 includes steps S410 to S470.

[0101] Step S410: Using the first input feature and the second input feature as the input layer, the liquid phase sintering theory as the hidden layer, and the deformation amount of the metal parts manufactured by binder spraying as the output layer, a neural network is constructed.

[0102] Step S420: Set the weights and bias values ​​of the input layer neurons.

[0103] Specifically, the weights and biases of the input layer neurons are set based on work experience.

[0104] Step S430: Based on the input layer data, the output layer data is obtained by performing forward propagation calculations on the neural network.

[0105] Step S440: The loss function is obtained by comparing the output layer data with the measured deformation value of the metal parts manufactured by adhesive spraying.

[0106] The loss function is actually the deformation prediction error, which is mainly used to verify the accuracy of the calculation method, the pre-set input layer neuron weights and bias values, and is also a reference for subsequent correction calculation methods.

[0107] Step S450: The gradient of each layer is obtained by backpropagation calculation of the neural network.

[0108] Step S460: Optimize the neural network parameters and minimize the error using the gradient descent method.

[0109] That is, after executing step S460, the parameters in the neural network (such as the weights and biases of neurons) and the loss function obtained in step S440 can be optimized to improve the confidence of the neural network.

[0110] Step S470: Return to the step of obtaining the output layer data until the neural network converges to obtain a preliminary deformation prediction model.

[0111] Thus, by executing steps S410 to S470, a preliminary deformation prediction model is created using a neural network algorithm. Steps S410 and S420 are executed to build the neural network. By executing steps S430 to S470, i.e., step S470 is to repeatedly execute steps S430 to S460, the neural network is repeatedly trained to complete the creation of the preliminary deformation prediction model and ensure that the preliminary deformation prediction model has a high confidence level.

[0112] Of course, in other embodiments, the preliminary deformation prediction model can also be created using deep learning algorithms, ensemble algorithms, association rule algorithms, reinforcement learning algorithms, etc.

[0113] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0114] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for deformation control in the manufacture of metal parts by adhesive spraying, characterized in that, Including the following steps: Based on experimental test data, a process database is obtained; the process database includes raw material characteristics, metal part dimensions, binder spraying process parameters, porosity characteristics of the metal part blank manufactured by binder spraying, and deformation amount of the metal part manufactured by binder spraying. Correlation analysis was performed on the different raw material characteristics, binder spraying process parameters and porosity characteristics of metal part blanks manufactured by binder spraying in the process database to screen out key factors affecting deformation as the first input feature; Correlation analysis was performed on the different metal component sizes, different porosity characteristics of the metal component blanks manufactured by binder spraying, and the deformation of the metal components manufactured by binder spraying in the process database to screen out the key factors affecting deformation as the second input features. Based on the liquid phase sintering theory, a preliminary deformation prediction model is created using the first input feature and the second input feature as input quantities and a machine learning algorithm. Using the Nash efficiency coefficient as a criterion, the preliminary deformation prediction model is iteratively optimized to obtain an optimized deformation prediction model.

2. The deformation control method for manufacturing metal parts by adhesive spraying according to claim 1, characterized in that, After obtaining the optimized deformation prediction model, the following steps are also included: Based on the dimensional data of the component to be manufactured, the deformation data of the component to be manufactured is obtained using the optimized deformation prediction model; Based on the deformation data of the part to be manufactured, a deformation inverse compensation design is performed on the printing model of the part to be manufactured to obtain a precision printing model.

3. The deformation control method for manufacturing metal parts by adhesive spraying according to claim 1, characterized in that, The characteristics of the raw materials include the particle size of the metal powder, the flowability of the metal powder, the morphology of the metal powder, the composition of the binder, the viscosity of the binder, and the surface tension of the binder.

4. The deformation control method for manufacturing metal parts by adhesive spraying according to claim 1, characterized in that, The adhesive spraying process parameters include printing layer thickness, powder spreading speed, adhesive saturation, and adhesive spraying amount per single layer.

5. The deformation control method for manufacturing metal parts by adhesive spraying according to claim 1, characterized in that, The porosity characteristics of the metal component blank manufactured by the binder spraying include porosity, pore space distribution, and pore size.

6. The deformation control method for manufacturing metal parts by adhesive spraying according to claim 1, characterized in that, The deformation of the metal parts manufactured by the adhesive spraying includes deformation data in the horizontal direction and deformation data in the vertical direction.

7. The deformation control method for manufacturing metal parts by adhesive spraying according to claim 1, characterized in that, In the process database, each parameter in the raw material properties has at least 5 data points that change in a stepwise manner, each parameter in the binder spraying process parameters has at least 5 data points that change in a stepwise manner, each parameter in the porosity characteristics of the metal part blank manufactured by binder spraying has at least 50 measured data points, and each parameter in the deformation amount of the metal part blank manufactured by binder spraying has at least 50 different data points.

8. The deformation control method for manufacturing metal parts by adhesive spraying according to claim 1, characterized in that, The steps for conducting correlation analysis on the characteristics of different raw materials, binder spraying process parameters, and porosity characteristics of metal part blanks manufactured by binder spraying, in order to obtain key parameters as the first input features, include the following steps: The raw material characteristics, binder spraying process parameters, and porosity characteristics of the metal part blank manufactured by binder spraying in the process database are each considered as the first set of independent and dependent variables. The first set of independent-dependent variables is divided into a first training set and a first validation set; Based on the data from the first training set, a model of the effect of independent variables on dependent variables was created and fitted using the stepwise regression method. In the aforementioned action model, by setting independent variables that have the least significant effect on the dependent variable at each step, the key factors affecting the deformation are obtained as the first input features. Using the independent variables in the first validation set as inputs, the dependent variable in the first validation set is predicted using the action model to obtain the prediction result; The mean squared error is calculated by comparing the measured results and the predicted results of the dependent variable in the first validation set. The interaction model is cross-validated using the mean squared error as the criterion.

9. The deformation control method for manufacturing metal parts by adhesive spraying according to claim 1, characterized in that, The steps of performing correlation analysis on the porosity characteristics of metal part blanks manufactured by binder spraying with different metal part sizes and deformation amounts to screen out key parameters as second input features include: The dimensions of the metal parts, the porosity characteristics of the metal part blanks manufactured by binder spraying, and the deformation of the metal parts manufactured by binder spraying in the process database are each considered as the second set of independent and dependent variables. The second group of independent-dependent variables is divided into a second training set and a second validation set; Based on the data from the second training set, a model of the effect of independent variables on dependent variables was created and fitted using stepwise regression. In the aforementioned action model, by setting independent variables that have the least significant effect on the dependent variable at each step, key factors influencing deformation are obtained as second input features. Using the independent variables in the second validation set as inputs, the dependent variable in the second validation set is predicted using the action model to obtain the prediction result; The mean squared error is calculated by comparing the measured and predicted results of the dependent variable in the second validation set. The interaction model is cross-validated using the mean squared error as the criterion.

10. The deformation control method for manufacturing metal parts by adhesive spraying according to claim 1, characterized in that, The steps for creating a preliminary deformation prediction model using machine learning algorithms include: Using the first input feature and the second input feature as the input layer, the liquid phase sintering theory as the hidden layer, and the deformation amount of the metal parts manufactured by binder spraying as the output layer, a neural network is constructed. Determine the weights and biases of the input layer neurons; Based on the input layer data, the output layer data is obtained by performing forward propagation calculations on the neural network; The loss function is obtained by comparing the output layer data with the measured values ​​of the deformation of the metal parts manufactured by adhesive spraying; The gradient of each layer is obtained by backpropagation calculation using the neural network. Optimize the neural network parameters and minimize the error using the gradient descent method; Return to the steps that yield the output layer data, and continue until the neural network converges to obtain a preliminary deformation prediction model.

Images