Manufacturing method and system for turbine blade of aero-engine

Abstract

The invention provides an aero-engine turbine blade manufacturing method and system. The method comprises the following steps: acquiring slurry attribute information of refractory slurry, and analyzing the slurry attribute information by using a first optimization model to obtain optimal slurry hanging parameters; obtaining a sanding attribute of a sanding nozzle, and analyzing the sanding attribute by using a second optimization model to obtain an optimal sanding parameter; and the optimal slurry coating parameters and the optimal sanding parameters are converted into optimal slurry coating execution parameters and optimal sanding execution parameters which can be recognized by the manipulator control system, slurry coating operation is executed on the wax pattern according to the optimal slurry coating execution parameters, and wax pattern sanding operation after slurry coating is executed according to optimal sanding. According to the invention, the automatic and precise operation of the shell making process is realized, the influence of human factors on the process quality is greatly reduced, the stability and consistency of the product quality are improved, and the production of high-quality and high-performance aero-engine turbine blades is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of turbine blade manufacturing, and in particular, to a manufacturing method and system for turbine blades of an aero-engine. Background Art

[0002] Turbine blades (manufactured from single crystals or directionally solidified alloys) are the core components of an aero-engine, located in the area with the highest temperature, the most complex stress, and the harshest environment, and are known as the "pearl on the crown". Their performance level, especially the temperature-bearing capacity, is an important indicator of the advancement of an engine model, and in a certain sense, it is also a remarkable indicator of the level of a country's aviation industry.

[0003] Investment casting is a key process in turbine blade manufacturing, and slurry hanging and sand spraying are the key links in this key process. In the shell-making process of investment casting, a manipulator will grab the wax module tree and immerse it in the refractory slurry at a specific order and speed, so that a layer of slurry is evenly covered on the surface of the wax mold, and then the sand spraying operation is carried out. The accuracy and stability of this series of actions are crucial for the uniformity and strength of the shell, and directly affect the final forming quality of the turbine blade. However, at present, specific parameters for slurry hanging and sand spraying are manually designed in advance for the manipulator to execute, but these specific parameters are obtained based on empirical values or theoretical values, and there are still large deficiencies in accuracy and need to be improved. Summary of the Invention

[0004] In order to solve the technical problems existing in the above background art, the present invention provides a manufacturing method, system, electronic device, computer storage medium, and computer program product for turbine blades of an aero-engine.

[0005] The present invention provides a manufacturing method for turbine blades of an aero-engine, and the method includes the following steps: Obtain the slurry property information of the refractory slurry, where the slurry property information includes refractory material properties and binder properties, the refractory material properties include the type and particle size of the refractory material, and the binder properties include the type, concentration, and content of the binder; Use a first optimization model to analyze the slurry property information to obtain the optimal slurry hanging parameters, where the optimal slurry hanging parameters include the optimal immersion speed, the optimal immersion depth, and the optimal residence time; Obtain the sand spraying properties of the sand spraying nozzle, where the sand spraying properties include the sand spraying speed, the type and particle size of the sand grains, and use a second optimization model to analyze the sand spraying properties to obtain the optimal sand spraying parameters, where the optimal sand spraying parameters include the optimal distance and the optimal angle; Convert the optimal slurry coating parameters and the optimal sand spraying parameters into the optimal slurry coating execution parameters and the optimal sand spraying execution parameters that can be recognized by the manipulator control system, and perform the slurry coating operation on the wax mold according to the optimal slurry coating execution parameters, and perform the sand spraying operation on the wax mold after slurry coating according to the optimal sand spraying.

[0006] Optionally, the first optimization model is constructed based on Transformer and includes an input embedding layer, an encoding layer, a decoding layer, and an output linear layer; wherein, the encoding layer includes multiple encoders; The input embedding layer is used to encode the refractory material attribute and the binder attribute in the slurry attribute information of the refractory slurry respectively to convert them into vector representations, and form the vectors into an input sequence vector; Each encoder in the encoding layer performs several downsampling convolution operations on the input sequence vector to extract features, fuse the extracted local features and global features, and then slice the fused features into first feature vectors of a preset size; The decoder is used to fuse each of the first feature vectors using a decoding attention network to obtain a second feature vector, and convert the second feature vector into the optimal slurry coating parameters.

[0007] Optionally, the preset size of the first feature vector is constructed in the following manner: Perform principal component analysis on the input slurry attribute information, and determine the appropriate preset size according to the variance ratio that can be explained.

[0008] Optionally, performing principal component analysis on the input slurry attribute information and determining the appropriate preset size according to the variance ratio that can be explained includes: Form the encoded slurry attribute information into a matrix, where each row represents a sample and each column represents a feature; Perform PCA decomposition on this matrix to obtain the principal components and the variance ratio they explain; Start accumulating the variance ratio explained by the largest principal component until a set threshold is reached, and the number of principal components used at this time is used as the preset size.

[0009] Optionally, the training data used by the first optimization model includes: refractory material attribute, binder attribute, immersion speed, immersion depth, residence time, first quality label; The training data used by the second optimization model includes: sand spraying speed, sand particle type, sand particle size, sand spraying distance, sand spraying angle, second quality label.

[0010] The present invention also provides an aero-engine turbine blade manufacturing system, which includes an acquisition module, a first optimization module, a second optimization module, and an execution module; The acquisition module is used to acquire the slurry property information of the refractory slurry. The slurry property information includes refractory material properties and binder properties. The refractory material properties include the type and particle size of the refractory material, and the binder properties include the type, concentration, and content of the binder. In addition, the sand spraying properties of the sand spraying nozzle are acquired. The sand spraying properties include the sand spraying speed, the type and particle size of the sand grains; The first optimization module is used to analyze the slurry property information using a first optimization model to obtain the optimal slurry hanging parameters. The optimal slurry hanging parameters include the optimal immersion speed, the optimal immersion depth, and the optimal residence time; The second optimization module is used to analyze the sand spraying properties using a second optimization model to obtain the optimal sand spraying parameters. The optimal sand spraying parameters include the optimal distance and the optimal angle; The execution module is used to convert the optimal slurry hanging parameters and the optimal sand spraying parameters into optimal slurry hanging execution parameters and optimal sand spraying execution parameters that can be recognized by the manipulator control system respectively, and perform the slurry hanging operation on the wax mold according to the optimal slurry hanging execution parameters, and perform the sand spraying operation on the wax mold after slurry hanging according to the optimal sand spraying.

[0011] Optionally, the first optimization model is constructed based on Transformer and includes an input embedding layer, an encoding layer, a decoding layer, and an output linear layer; wherein, the encoding layer includes multiple encoders; The input embedding layer is used to encode the refractory material properties and the binder properties in the slurry property information of the refractory slurry respectively to convert them into vector representations, and form the input sequence vectors by combining the vectors; Each encoder in the encoding layer performs several downsampling convolution operations on the input sequence vectors to extract features, fuses the extracted local features and all features, and then cuts the fused features into first feature vectors of a preset size; The decoder is used to fuse and process each of the first feature vectors using a decoding attention network to obtain a second feature vector, and convert the second feature vector into the optimal slurry hanging parameters.

[0012] Optionally, the preset size of the first feature vector is constructed by the following method: Perform principal component analysis on the input slurry property information, and determine the appropriate preset size according to the variance ratio that can be explained.

[0013] Optionally, performing principal component analysis on the input slurry property information and determining the appropriate preset size according to the variance ratio that can be explained includes: Forming the encoded slurry property information into a matrix, where each row represents a sample and each column represents a feature; Performing PCA decomposition on the matrix to obtain the principal components and the variance ratio they explain; Starting from the largest principal component, accumulating the variance ratio it explains until a set threshold is reached, and using the number of principal components used at this time as the preset size.

[0014] Optionally, the training data used by the first optimization model includes: refractory material properties, binder properties, immersion speed, immersion depth, residence time, and first quality labels; The training data used by the second optimization model includes: sand spraying speed, sand particle type, sand particle size, sand spraying distance, sand spraying angle, and second quality labels.

[0015] The present invention provides an electronic device, including: a memory storing executable program code; a processor coupled to the memory; the processor calling the executable program code stored in the memory and executing the method described in any one of the above.

[0016] The present invention also provides a computer storage medium, on which a computer program is stored, and when the computer program is run by a processor, it executes the method described in any one of the above.

[0017] The present invention also provides a computer program product, which includes a computer program stored in a computer storage medium, and when the computer program is executed by a processor of an electronic device, it implements the method described in any one of the above.

[0018] The beneficial effects of the present invention are at least as follows: The method for manufacturing an aeroengine turbine blade of the present invention comprehensively obtains relevant property information of refractory slurry and sand spraying nozzles, and uses an optimization model for in-depth analysis to accurately obtain the optimal slurry hanging parameters and the optimal sand spraying parameters, realizing the automated and precise operation of the shell-making process, greatly reducing the influence of human factors on the process quality, improving the stability and consistency of product quality, contributing to the production of high-quality and high-performance aeroengine turbine blades, meeting the stringent requirements of the aviation field for blade manufacturing, and at the same time improving production efficiency and reducing production costs. Description of the Drawings

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0020] Figure 1 is a schematic flowchart of a method for manufacturing an aeroengine turbine blade disclosed in an embodiment of the present invention; Figure 2 is a schematic structural diagram of a first optimization model disclosed in an embodiment of the present invention; Figure 3 is a schematic structural diagram of an aeroengine turbine blade manufacturing system disclosed in an embodiment of the present invention. Detailed implementation manners

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0022] For the above technical problems, please refer to Figure 1 , an embodiment of the present invention discloses a method for manufacturing an aeroengine turbine blade, and the method includes the following steps: Obtain the slurry property information of the refractory slurry, where the slurry property information includes refractory material properties and binder properties. The refractory material properties include the type and particle size of the refractory material, and the binder properties include the type, concentration, and content of the binder; Use the first optimization model to analyze the slurry property information to obtain the optimal slurry hanging parameters, where the optimal slurry hanging parameters include the optimal immersion speed, the optimal immersion depth, and the optimal residence time; Obtain the sand spraying property of the sand spraying nozzle, where the sand spraying property includes the sand spraying speed, the type and particle size of the sand grains. Use the second optimization model to analyze the sand spraying property to obtain the optimal sand spraying parameters, where the optimal sand spraying parameters include the optimal distance and the optimal angle; Convert the optimal slurry hanging parameters and the optimal sand spraying parameters into the optimal slurry hanging execution parameters and the optimal sand spraying execution parameters that can be recognized by the manipulator control system, and perform the slurry hanging operation on the wax mold according to the optimal slurry hanging execution parameters, and perform the sand spraying operation on the wax mold after slurry hanging according to the optimal sand spraying execution parameters.

[0023] The present invention pre - constructs a first optimization model and a second optimization model, which are respectively used to predict the optimal sizing parameters and sand - spraying parameters. Then, the optimal sizing parameters and sand - spraying parameters are converted into optimal sizing execution parameters and optimal sand - spraying execution parameters that can be recognized by the manipulator control system. The manipulator control system controls the manipulator to immerse the wax mold into the refractory slurry at the optimal immersion speed and optimal immersion depth, and keep it in the refractory slurry for the optimal residence time, so that the actual sizing thickness of the wax mold meets the expected designed sizing thickness; then it controls the sized wax mold to approach the sand - spraying nozzle at the optimal angle and optimal distance, so that the sand grain thickness in each area of the wax mold is consistent. The shell structure of the aero - engine turbine blade cast by using the above - mentioned wax mold is denser, the strength distribution is uniform, and the blade forming quality is higher.

[0024] The manufacturing method of the aero - engine turbine blade of the present invention comprehensively obtains the relevant attribute information of the refractory slurry and the sand - spraying nozzle, and uses the optimization model for in - depth analysis to accurately obtain the optimal sizing parameters and optimal sand - spraying parameters, realizing the automated and precise operation of the shell - making process, greatly reducing the influence of human factors on the process quality, improving the stability and consistency of the product quality, contributing to the production of high - quality and high - performance aero - engine turbine blades, meeting the stringent requirements of the aviation field for blade manufacturing, while improving production efficiency and reducing production costs.

[0025] Optionally, the first optimization model is constructed based on Transformer and includes an input embedding layer, an encoding layer, a decoding layer, and an output linear layer; among them, the encoding layer includes multiple encoders; The input embedding layer is used to encode the refractory material attribute and the binder attribute in the slurry attribute information of the refractory slurry respectively to convert them into vector representations, and form the input sequence vectors by combining the vectors. Each encoder in the encoding layer performs several down - sampling convolution operations on the input sequence vectors to extract features, fuses the extracted local features and global features, and then slices the fused features into first feature vectors of a preset size. The decoder is used to fuse - process each of the first feature vectors using a decoding attention network to obtain a second feature vector, and convert the second feature vector into the optimal sizing parameters.

[0026] In this example, the first optimization model is constructed based on the Transformer architecture and mainly consists of an input embedding layer, an encoding layer, a decoding layer, and an output linear layer. This architecture aims to process the attribute information of the refractory slurry and, through a complex feature extraction and information fusion mechanism, finally output the optimal sizing parameters, that is, the optimal immersion speed, optimal immersion depth, and optimal residence time.

[0027] Input Embedding Layer: The main task of this layer is to encode the slurry property information of the input refractory slurry. The property information of the refractory slurry includes refractory material properties (i.e., refractory material type, particle size) and binder properties (i.e., binder type, concentration, and content), and these properties are usually discrete or data with different dimensions. By encoding these different types of property information separately and converting them into a unified vector representation, the model can perform mathematical processing on this information. Finally, the encoded vector groups of different properties are combined into an input sequence vector to provide input for the subsequent encoding layer. For example, if the refractory material type is encoded as a vector v1, the refractory material particle size is encoded as vector v2, and the binder type, concentration, and content are encoded as vectors v3, v4, and v5 respectively, then the input sequence vector can be represented as [v1, v2, v3, v4, v5].

[0028] Encoding Layer: It contains multiple encoders, which are used to perform in-depth feature extraction and information processing on the input sequence vector. Each encoder can be regarded as an information processing unit, and the stacking of multiple encoders enables the model to learn feature representations at different levels.

[0029] Perform several downsampling convolution operations on the input sequence vector to extract local features. The downsampling convolution operation performs convolution calculations on the input sequence by sliding the convolution kernel, which can effectively reduce the data dimension while capturing local features. For example, in image processing, downsampling convolution can convert a high-resolution image into a low-resolution image while retaining the main features of the image. Here, a similar process is applied to the dimension and information volume of the input sequence vector. The downsampling convolution operation helps reduce the computational amount while capturing the local correlations between different properties and extracting local features.

[0030] Fuse the local features extracted by the downsampling convolution with the global features. This setting is to comprehensively consider local information and global information, avoiding losing global information by only focusing on local information or ignoring local details by only relying on global information. This fusion can be achieved in various ways, such as concatenation, addition, or weighted addition, enabling the model to comprehensively consider feature information at different levels and scopes.

[0031] To divide the features into multiple parts according to a certain logic or rule for the convenience of subsequent processing by the decoding layer, the fused features are also sliced into first feature vectors of a preset size. Each of the sliced first feature vectors contains feature information in different aspects, providing more targeted data for subsequent decoding.

[0032] Decoding layer: The first feature vector is fused using a decoding attention network. The attention network is a core part of Transformer, allowing the model to focus on different parts of the input, thereby selectively attending to important information. The attention network calculates the correlation weights between the first feature vectors, sums them up with weights, and generates a new feature representation, i.e., the second feature vector.

[0033] Output linear layer: This layer is used to convert the second feature vector output by the decoding layer into the optimal sizing parameters. Through linear transformation, the second feature vector is mapped to the specific sizing parameter space. For example, the linear transformation formula y = Wx + b is used, where x is the second feature vector, W is the weight matrix, b is the bias vector, and y is the finally output optimal sizing parameter.

[0034] Optionally, the second optimization model is the same as the first optimization model.

[0035] In this example, the second optimization model for determining the optimal sanding parameters can adopt the same model structure as the first optimization model, that is, it is also constructed based on Transformer. The specific model structure and the functions of each layer are not elaborated here.

[0036] Optionally, the preset size of the first feature vector is constructed in the following manner: Perform principal component analysis on the input slurry property information, and determine the appropriate preset size according to the proportion of variance that can be explained.

[0037] In this example, the present invention can flexibly determine the preset size in the first optimization model according to specific situations, to balance the performance of the model, avoid overfitting, and ensure that the input slurry property information can be fully extracted and characterized, providing more effective support for the optimization of the sizing parameters in the manufacture of turbine blades. For example, the diversity and complexity of the slurry property information of the input refractory slurry (such as the type and particle size of the refractory material, the type, concentration and content of the binder) (in addition to single refractory materials and single binders in the refractory slurry, there may also be multiple refractory materials and multiple binders) will affect the size of the required feature vector. If the input property types are numerous and the range of property values is wide, a larger feature vector may be required to fully characterize this information.

[0038] The present invention specifically adopts a data-driven method, performs principal component analysis on the input slurry property information, and determines the appropriate preset size according to the proportion of variance that can be explained. A threshold can be set (such as the proportion of explained variance reaching 95%), and according to the PCA analysis results, the number of principal components required to reach this threshold is selected as a reference for the preset size.

[0039] Optionally, performing principal component analysis on the input slurry property information and determining the appropriate preset size according to the proportion of variance that can be explained, including: Forming the encoded slurry property information into a matrix, where each row represents a sample and each column represents a feature; Performing PCA decomposition on this matrix to obtain the principal components and the proportion of variance they explain; Starting from the largest principal component, accumulating the proportion of variance it explains until a set threshold is reached, and using the number of principal components used at this time as the preset size.

[0040] Optionally, the training data used by the first optimization model includes: refractory material properties, binder properties, immersion speed, immersion depth, residence time, first quality label; The training data used by the second optimization model includes: sand spraying speed, sand grain type, sand grain size, sand spraying distance, sand spraying angle, second quality label.

[0041] In this example, the first optimization model and the second optimization model are respectively trained with the above two types of training data, so as to obtain models that can accurately predict the optimal slurry hanging parameters and the optimal sand spraying parameters. This training can adopt conventional local training or a distributed training method, and the present invention does not make specific limitations on this.

[0042] In addition, a suitable loss function is selected during the training process, such as mean square error (MSE) or mean absolute error (MAE), to measure the difference between the model prediction value and the true value; and, preferably Adam or SGD is selected as the optimizer to update the parameters of the model to minimize the loss function.

[0043] The embodiment of the present invention also provides an aeroengine turbine blade manufacturing system, and the system includes an acquisition module 10, a first optimization module 12, a second optimization module 14, and an execution module 16; The acquisition module 10 is used to acquire the slurry property information of the refractory slurry, and the slurry property information includes refractory material properties and binder properties. The refractory material properties include refractory material type and particle size, and the binder properties include binder type, concentration and content; and, acquire the sand spraying properties of the sand spraying nozzle, and the sand spraying properties include sand spraying speed, sand grain type and particle size; The first optimization module 12 is used to analyze the slurry property information using the first optimization model to obtain the optimal slurry hanging parameters, and the optimal slurry hanging parameters include the optimal immersion speed, the optimal immersion depth and the optimal residence time; The second optimization module 14 is configured to analyze the sand spraying attribute using a second optimization model to obtain optimal sand spraying parameters, where the optimal sand spraying parameters include an optimal distance and an optimal angle; The execution module 16 is configured to convert the optimal slurry hanging parameters and the optimal sand spraying parameters into optimal slurry hanging execution parameters and optimal sand spraying execution parameters that can be recognized by the manipulator control system, and perform a slurry hanging operation on the wax mold according to the optimal slurry hanging execution parameters, and perform a sand spraying operation on the wax mold after slurry hanging according to the optimal sand spraying.

[0044] An embodiment of the present invention further provides an electronic device, including: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory and executes the method according to any one of the above embodiments.

[0045] An embodiment of the present invention further provides a computer storage medium, on which a computer program is stored, and when the computer program is run by a processor, it executes the method according to any one of the above embodiments.

[0046] An embodiment of the present invention further provides a computer program product, where the computer program product includes a computer program stored in a computer storage medium, and when the computer program is executed by a processor of an electronic device, it implements the method according to any one of the above embodiments.

[0047] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (devices), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, and the combination of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0048] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0049] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, thereby the instructions executed on the computer or other programmable apparatus provide steps for realizing the functions specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 steps for realizing the functions specified in one block or a plurality of blocks.

[0050] As described above, only the specific embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A manufacturing method for an aero-engine turbine blade, characterized in that: The method includes the following steps: Obtain the slurry property information of the refractory slurry, where the slurry property information includes refractory material properties and binder properties. The refractory material properties include the type and particle size of the refractory material, and the binder properties include the type, concentration, and content of the binder; Use the first optimization model to analyze the slurry property information to obtain the optimal slurry hanging parameters, where the optimal slurry hanging parameters include the optimal immersion speed, optimal immersion depth, and optimal residence time; Obtain the sand spraying properties of the sand spraying nozzle, where the sand spraying properties include the sand spraying speed, sand particle type, and particle size. Use the second optimization model to analyze the sand spraying properties to obtain the optimal sand spraying parameters, where the optimal sand spraying parameters include the optimal distance and optimal angle; Respectively convert the optimal slurry hanging parameters and the optimal sand spraying parameters into optimal slurry hanging execution parameters and optimal sand spraying execution parameters that can be recognized by the manipulator control system, and perform the slurry hanging operation on the wax mold according to the optimal slurry hanging execution parameters, and perform the sand spraying operation on the wax mold after slurry hanging according to the optimal sand spraying.

2. A manufacturing method of an aeroengine turbine blade according to claim 1, characterized in that: The first optimization model is constructed based on Transformer and includes an input embedding layer, an encoding layer, a decoding layer, and an output linear layer; among them, the encoding layer includes multiple encoders; The input embedding layer is used to encode the refractory material properties and the binder properties in the slurry property information of the refractory slurry respectively to convert them into vector representations, and form the input sequence vectors by combining the vectors; Each encoder in the encoding layer performs several downsampling convolution operations on the input sequence vectors to extract features, fuse the extracted local features and all features, and then divide the fused features into first feature vectors of a preset size; The decoder is used to fuse and process each of the first feature vectors using a decoding attention network to obtain second feature vectors, and convert the second feature vectors into the optimal slurry hanging parameters.

3. A manufacturing method of an aeroengine turbine blade according to claim 2, characterized in that: The preset size of the first feature vector is constructed in the following manner: Perform principal component analysis on the input slurry property information, and determine the appropriate preset size according to the variance ratio that can be explained.

4. A method for manufacturing an aeroengine turbine blade according to claim 3, characterized in that: The performing principal component analysis on the input slurry property information and determining the appropriate preset size according to the variance ratio that can be explained includes: Form a matrix with the encoded slurry property information, where each row represents a sample and each column represents a feature; Perform PCA decomposition on this matrix to obtain the principal components and the variance ratio they explain; Start accumulating the variance ratio explained by the largest principal component until the set threshold is reached, and use the number of principal components used at this time as the preset size.

5. A method for manufacturing an aeroengine turbine blade according to claim 1, characterized in that: The training data used by the first optimization model includes: refractory material properties, binder properties, immersion speed, immersion depth, residence time, first quality label; The training data used by the second optimization model includes: sand spraying speed, sand particle type, sand particle size, sand spraying distance, sand spraying angle, second quality label.

6. An aero-engine turbine blade manufacturing system, characterized in that: The system includes an acquisition module, a first optimization module, a second optimization module, and an execution module; The acquisition module is configured to acquire the slurry property information of the refractory slurry. The slurry property information includes refractory material properties and binder properties. The refractory material properties include the type of refractory material and the particle size. The binder properties include the type of binder, the concentration, and the content. In addition, the sand spraying properties of the sand spraying nozzle are acquired. The sand spraying properties include the sand spraying speed, the type of sand grains, and the particle size. The first optimization module is configured to analyze the slurry property information using a first optimization model to obtain the optimal slurry hanging parameters. The optimal slurry hanging parameters include the optimal immersion speed, the optimal immersion depth, and the optimal residence time. The second optimization module is configured to analyze the sand spraying properties using a second optimization model to obtain the optimal sand spraying parameters. The optimal sand spraying parameters include the optimal distance and the optimal angle. The execution module is configured to convert the optimal slurry hanging parameters and the optimal sand spraying parameters into optimal slurry hanging execution parameters and optimal sand spraying execution parameters that can be recognized by the manipulator control system respectively, and perform the slurry hanging operation on the wax mold according to the optimal slurry hanging execution parameters, and perform the sand spraying operation on the wax mold after slurry hanging according to the optimal sand spraying execution parameters.

7. An aero-engine turbine blade manufacturing system according to claim 6, characterized in that: The first optimization model is constructed based on Transformer and includes an input embedding layer, an encoding layer, a decoding layer, and an output linear layer. Among them, the encoding layer includes multiple encoders. The input embedding layer is configured to encode the refractory material properties and the binder properties in the slurry property information of the refractory slurry respectively to convert them into vector representations, and form the vector groups into an input sequence vector. Each of the encoders in the encoding layer performs several downsampling convolution operations on the input sequence vector to extract features, fuses the extracted local features and global features, and then slices the fused features into first feature vectors of a preset size. The decoder is configured to fuse and process each of the first feature vectors using a decoding attention network to obtain a second feature vector, and convert the second feature vector into the optimal slurry hanging parameters.

8. An electronic device, comprising: A memory storing executable program code; A processor coupled to the memory; characterized in that: the processor calls the executable program code stored in the memory and executes the method according to any one of claims 1-5.

9. A computer storage medium, on which a computer program is stored, characterized in that: When the computer program is run by a processor, it executes the method according to any one of claims 1-5.

10. A computer program product, comprising a computer program stored in a computer storage medium, characterized in that: When the computer program is executed by the processor of an electronic device, it implements the method according to any one of claims 1-5.

Images