A real-time temperature field prediction method for turbine blade arc additive
By establishing a simulation database and BP neural network model in arc additive manufacturing, the problem of real-time temperature field prediction of turbine blade arc additive in the prior art is solved, and accurate prediction of turbine blade additive temperature and process parameters are achieved. The optimization of arc additives improves the formation accuracy and efficiency.
Patent Information
- Application Number
- CN202211330779.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-10-28
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2042-10-28
AI Technical Summary
The prior art is difficult to predict the turbine blade temperature field in arc additive manufacturing in real time, resulting in low field control efficiency and lack of real-time temperature field modeling under the action of heat sources of different parameters.
By establishing a temperature field of simulated turbine blade additive process based on the profile feature size, establishing a welding simulation database, building a welding data platform, and using BP neural network to establish a temperature field prediction model, conducting training and verification, and achieving arc increase of turbine blades Prediction of real-time temperature field of materials.
Accurate prediction of real-time temperature of turbine blade additives is achieved, reflecting the evolution law of the thermal process, helping to select reasonable process parameters and thermal input parameters, improving the accuracy and efficiency of arc additive forming, and reducing costs.
Smart Images

Figure CN115563736B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of temperature field prediction, and in particular to a method for predicting the real-time temperature field of turbine blade arc additive. Background Art
[0002] At present, arc additive manufacturing technology has been widely used in high-temperature alloy welding, including aerospace, vehicles, ships and other industries. Arc additive technology is a rapid metal additive method that forms from point to surface and then accumulates layer by layer. The arc, as a heat source, generates huge heat to quickly melt the metal wire and solidify it on the substrate. This is a reliable solid-phase connection technology, but precisely because of the continuous input of heat source during the arc additive process, the temperature inside the formed part continues to accumulate and evolve. Studying the generation and change of the temperature field in the additive workpiece is very important for understanding the arc additive manufacturing process.
[0003] Chinese Patent 201810973005.0 discloses a thermal field calculation method for arc additive manufacturing of complex feature structural parts. This method establishes a finite element model for thermal field calculation based on a generalized double ellipsoid heat source to perform thermal field calculation of complex feature components; Chinese Patent 202110474227.X proposes a heat source numerical simulation method for laser additive manufacturing process, laying the foundation for the numerical simulation analysis of process temperature field in laser additive manufacturing. Chinese Patent 202110871691.2 proposes an analysis method for the temperature field of gas turbine blades, which extracts the manufacturing parameter data of the turbine blades and refers to the target map to estimate the casting temperature field. The software simulation used in the above-mentioned prior art has great deficiencies, the preliminary preparation is long and the server needs to be used for long-term calculation, which is difficult to meet the needs of real-time control on site. In addition, some existing technologies only analyze the establishment and optimization of heat source models, and some only establish temperature field models for laser additive manufacturing. There are few comprehensive analyses of the real-time temperature field of arc additive manufacturing under the action of heat sources with different parameters for turbine blades, and the prediction of the real-time temperature field of arc additive manufacturing of turbine blades is not involved. Summary of the invention
[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.
[0005] In view of the above existing problems, the present invention is proposed.
[0006] Therefore, the present invention provides a real-time temperature field prediction method for turbine blade arc additive, which can solve the problem that the temperature field model of the traditional analysis method has low calculation efficiency and cannot meet the real-time control requirements on site.
[0007] In order to solve the above technical problems, the present invention provides the following technical solution, a method for predicting the real-time temperature field of turbine blade arc additive, comprising:
[0008] Establish the temperature field of the turbine blade additive process based on the characteristic dimensions of the profile;
[0009] Establishing a welding simulation database based on the temperature field, and storing the temperature field calculation results as simulation samples in the database;
[0010] Build a welding data platform based on welding simulation database and actual welding data;
[0011] Establishing a temperature field prediction model of a BP neural network based on the welding data platform;
[0012] The established temperature field prediction model is trained and verified.
[0013] As a preferred solution of the real-time temperature field prediction method for turbine blade arc additive described in the present invention, the process of establishing a temperature field simulation process for turbine blade additive includes establishing a model and defining material properties, dividing units and grids, adding initial and boundary conditions, applying a moving heat source, setting a time step and solving and post-processing the results.
[0014] As a preferred method of predicting the real-time temperature field of turbine blade arc material addition according to the present invention,
[0015] Case, wherein: the simulated samples are obtained by designing a rotation regression test combination.
[0016] As a preferred solution of the method for real-time temperature field prediction of turbine blade arc additive described in the present invention, the data processing of the sample is to use the mapminmax function in Matlab to normalize the imported training set and test set.
[0017] As a preferred solution of the method for predicting the real-time temperature field of turbine blade arc additive according to the present invention, the specific calculation formula of the mapminmax function is expressed as:
[0018]
[0019] Where x represents the input value or expected output value in the test sample data; x min Indicates the minimum value in the sample; x maxRepresents the maximum value in the sample; y represents the normalized value in the range [-1,1], y min =-1;y max =1.
[0020] As a preferred solution of the method for predicting the real-time temperature field of turbine blade arc additive described in the present invention, the welding data platform includes a material library, an experimental library, a simulation library and a prediction calculation module.
[0021] As a preferred solution of the method for real-time temperature field prediction of turbine blade arc additive described in the present invention, the BP neural network is a multi-layer network for error back propagation, which forward calculates the error of each previous layer through the output layer error.
[0022] As a preferred solution of the method for predicting the real-time temperature field of turbine blade arc additive according to the present invention, the working process expression of the neuron in the BP neural network is expressed as follows:
[0023]
[0024] in, represents m input variables, is the connection weight, represents summation, b is the bias value, f is the activation function, and y is the output variable.
[0025] As a preferred solution of the real-time temperature field prediction method for arc additive manufacturing of turbine blades described in the present invention, the BP neural network input variables are blade root form, blade crown form, layer thickness, inlet and outlet steam edge radius, cross-sectional position and curve parameters, and the output variable is temperature.
[0026] As a preferred solution of the real-time temperature field prediction method for arc additive of turbine blades described in the present invention, the training and verification of the temperature field prediction model includes:
[0027] Data construction: The data obtained from several sets of blade simulation analysis are stored in the database through the server;
[0028] Data formatting: Reorganize the data by dividing it into two independent data files, one for training and the other for testing, and ensure that they are consistent with the neural network input format;
[0029] Convergence judgment: When the system converges satisfactorily and the error threshold is met, the ANN parameters are stored for testing. When the system does not converge, the training parameters must be modified until satisfactory results are obtained.
[0030] Find the optimal solution: Test the solutions with a data test file and use the solution with the smallest error in the prediction process during background processing.
[0031] Beneficial effects of the present invention: The present invention proposes a method for predicting the real-time temperature field of turbine blade arc additive. After training the BP neural network with a large number of samples, the real-time temperature of turbine blade additive can be predicted. The established BP neural network model can reflect the evolution law of the thermal process of blade additive, so as to select reasonable process parameters and heat input parameters, that is, input model and layer thickness, and quickly calculate the real-time corresponding arc heat input power of each layer forming path, so as to provide guarantee for the precise forming of arc additive. Combining welding technology, database technology and neural network modeling can more accurately predict the additive temperature under certain processing conditions, provide a theoretical basis for reducing heat loss, effectively reduce costs and improve efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:
[0033] Figure 1 An overall flow chart of a method for predicting real-time temperature field of arc material addition for turbine blades provided by one embodiment of the present invention;
[0034] Figure 2 A flow chart of establishing a method for predicting the real-time temperature field of a turbine blade arc material addition process and simulating the temperature field of a turbine blade material addition process provided by an embodiment of the present invention;
[0035] Figure 3 A neuron workflow diagram of a method for predicting real-time temperature field of arc additive for turbine blades provided by one embodiment of the present invention;
[0036] Figure 4 A BP neural network structure diagram of a method for predicting the real-time temperature field of arc additive for turbine blades provided by one embodiment of the present invention;
[0037] Figure 5 A BP neural network model training flow chart of a method for predicting the real-time temperature field of arc additive for turbine blades provided by one embodiment of the present invention;
[0038] Figure 6 A comparison diagram of the errors of different numbers of neurons in hidden layers of a method for predicting the real-time temperature field of arc additive for turbine blades provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0039] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.
[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0041] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.
[0042] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.
[0043] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.
[0044] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0045] Example 1
[0046] Reference Figure 1-5, which is the first embodiment of the present invention, provides a method for predicting the real-time temperature field of arc additive for turbine blades, comprising:
[0047] S1: Establish the temperature field of the additive process of the turbine blade based on the characteristic dimensions of the profile;
[0048] Furthermore, when the finite element software Ansys Workbench simulates the temperature field of the turbine blade additive process, it must first establish a turbine blade structure model based on the size of the required turbine blade model, set the layer thickness and material properties of the turbine blade, and then divide the units and networks according to the required accuracy. Considering the influencing factors, it is necessary to first determine the initial conditions and boundary conditions, then add a mobile heat source to the model, and finally set the time step to achieve the final solution.
[0049] It should be noted that there are many different types of turbine blades and many different parameters of mobile heat sources, but this patent invents a general method. After the model training and verification, the additive temperature of the turbine blades of the required model under any parameters of the mobile heat source can be predicted.
[0050] It should also be noted that this patent focuses on blades in order to reduce the difficulty of the experiment and increase the feasibility of the method for other similar structural parts. The ultimate goal of this method is not to limit the additive object.
[0051] S2: Establish a welding simulation database based on the temperature field, and store the temperature field calculation results as simulation samples in the database;
[0052] Furthermore, the samples were obtained by designing a rotation regression test combination and conducting 40 sets of turbine blade additive temperature field simulation tests under the same test conditions and within the selected process parameter range. The obtained data was divided into 3:1, 30 sets of data were used as the training set of the network, and 10 sets of data were used as the test set of the network. The experimental data were normalized before network training and testing.
[0053] It should be noted that the advantage of the rotational regression test combination is that the predicted values of each treatment combination can be directly compared, so as to find out the areas with relatively good predicted values.
[0054] It should also be noted that normalization can make input variables with different dimensions have the same importance after processing.
[0055] Furthermore, the mapminmax function in Matlab is used to normalize the imported training set and test set so that the output data range is fixed between [-1, 1].
[0056] It should be noted that the specific calculation formula of the mapminmax function is expressed as:
[0057]
[0058] Where x represents the input value or expected output value in the test sample data; x min Indicates the minimum value in the sample; x max Represents the maximum value in the sample; y represents the normalized value in the range [-1,1], y min =-1;y max =1.
[0059] S3: Build a welding data platform based on the welding simulation database and actual welding data;
[0060] Furthermore, the temperature field data required for the welding temperature field prediction process is collected in the welding process prediction data platform. The database uses MS SQL Server, and the data platform architecture adopts the B / S (browser / server) mode. Based on actual welding and numerical simulation, a digital information platform is established, and on this basis, the prediction research on the temperature field of the weld is carried out, and compared with the actual welding experiment, the accuracy of the prediction results is verified.
[0061] It should be noted that the actual welding data is obtained by using a pyrometer or an infrared thermal imager to capture the additive welding process in real time and record the welding temperature field information data.
[0062] It should also be noted that the actual welding data is used in conjunction with the simulation data. For every set of simulation data, there must be a corresponding set of actual welding data.
[0063] S4: Establish a temperature field prediction model of BP neural network based on welding data platform;
[0064] Furthermore, a BP neural network is used and a structural design is performed to determine the input variables, output variables and the number of hidden layers, and a temperature field prediction model based on the BP neural network is established.
[0065] It should be noted that the BP neural network is a multi-layer network with error back propagation. The error of each previous layer is calculated forward through the error of the output layer. That is, the network learning process is carried out in the forward direction and the error transmission is carried out in the reverse direction.
[0066] Furthermore, during the training process, the input value is weighted and reaches the neuron threshold before being passed to the next neuron through the activation function. When the output error is greater than the set error value, the error backpropagation correction process will be performed.
[0067] It should be noted that the working process of neurons can be expressed by functional expressions, as follows:
[0068]
[0069] in, represents m input variables, is the connection weight, represents summation, b is the bias value, f is the activation function, and y is the output variable.
[0070] Furthermore, during the forward propagation of the signal in the neural network, the input sample is passed from the input layer, processed layer by layer by the hidden layer, and then passed to the output layer.
[0071] It should be noted that currently the determination of the hidden layer is mainly estimated using empirical formulas.
[0072]
[0073] Where y represents the number of neurons in the hidden layer; a represents the number of variables in the input layer; b represents the number of variables in the output layer; c is a constant between [1,10], and c is an adjustment constant, which is a range of values selected based on experience and the following basic principles.
[0074] It should also be noted that the number of neurons in the hidden layer determines the network's ability to process nonlinear functions.
[0075] Furthermore, the blade root form, blade crown form, layer thickness, inlet and outlet steam edge radius, cross-sectional position and curve parameters are used as input variables of the neural network, and temperature is used as the output variable. The temperature field information of the corresponding position is obtained according to the grid node coordinates of the turbine blade in the established three-dimensional model. Combined with the value range of c, the range of the number of hidden layers can be obtained as [4,13]. In order to determine the specific number of neurons in the hidden layer, the values within the range are substituted into the network for learning and prediction.
[0076] It should be noted that for blades of different models, the characteristic dimensions of the profile, including the inlet and outlet radius, chord width, etc., are extracted to determine the input variables and layer thickness. The basic principle for determining the number of hidden layer nodes is: to take the most compact structure possible, that is, to take the least number of hidden layer nodes possible, while meeting the accuracy.
[0077] Furthermore, after the hidden layer is determined, the hyperbolic tangent sigmoid function is selected as the activation function in the BP neural network learning process. The number of network iterations is set to 100 times, and the network learning rate is set to 0.01.
[0078] It should be noted that Sigmoid is a commonly used activation function, which can convert the weighted sum into a value between 0 and 1, and is used to convert the logistic regression value into a binary value, and then perform category division. The introduction of the activation function can solve nonlinear problems, and the introduced nonlinear function is used to manage each hidden layer node.
[0079] It should also be noted that the difference between the learning result and the sample is defined as the loss value. While avoiding overfitting, the number of iterations is determined by the test effect, but the iteration is generally terminated when the loss value converges. The traditional default values of the learning rate are 0.1 and 0.01. The default value of 0.01 is usually suitable for multi-layer neural networks.
[0080] S5: Train and verify the established temperature field prediction model.
[0081] Furthermore, the data is constructed by storing the data obtained from the simulation analysis of several sets of blades in a database through a server;
[0082] Furthermore, the data was formatted to be consistent with the input format of the neural network and reorganized by splitting the data into two separate data files, one for training and validation and the other for testing.
[0083] Furthermore, if the system converges satisfactorily to confirm that the error threshold is met, the ANN parameters are stored for testing. If the system does not converge, the training parameters must be modified until they are satisfactory. Since the system can converge to different solutions with different initial conditions and training parameters, the above process is repeated to obtain a large number of solutions, which can increase the possibility of finding the overall optimal solution.
[0084] It should be noted that the ANN parameters represent the relationship weights and the critical values of the units.
[0085] Furthermore, the solutions are tested using a data test file and the solution with the minimum error is used in the prediction process during background processing.
[0086] It should be noted that the process of learning and training a neural network is actually a function approximation process, that is, a process of continuously correcting the connection weights of the interactions between neurons within the neural network.
[0087] Furthermore, during the training process, the weights and thresholds of the network are repeatedly adjusted to reduce the mean square error between the network output and the target output. The test samples are used to evaluate the performance of the trained network. The validation samples are used to verify the predicted values and the measured values to verify the accuracy of the network training.
[0088] Furthermore, by inputting the model and layer thickness, the arc heat input power corresponding to each layer of the forming path can be quickly calculated in real time, providing a guarantee for accurate arc additive forming.
[0089] It should be noted that the purpose of rapid temperature field calculation is to calculate the temperature at any point in each layer and each path of the additive process under the condition of known input thermal power. Correspondingly, when the additive forming temperature requirement has been determined, the input arc power can be obtained by the prediction model, thereby providing real-time control parameters for precise arc additive forming.
[0090] Example 2
[0091] Reference Figure 6 , which is an embodiment of the present invention, provides a method for predicting the real-time temperature field of arc additive for turbine blades. The prediction mean square error of the learning sample and the verification sample is used as the inspection standard to obtain the training error of the network under different hidden layer neurons. The error values are shown in Table 1:
[0092] Table 1 Comparison of errors in the number of neurons in different hidden layers
[0093] Simulation Number Network structure Learning sample mean square error (mm) Verification sample mean square error (mm) 1 6-4-1 0.10 0.05 2 6-4-1 0.03 0.03 3 6-4-1 0.02 0.05 4 6-4-1 0.018 0.10 5 6-4-1 0.017 0.08 6 6-4-1 0.016 0.07 7 6-4-1 0.015 0.06 8 6-4-1 0.015 0.04 9 6-4-1 0.015 0.04 10 6-4-1 0.015 0.04
[0094] From Table 1, we can get the trend of error change by visually analyzing the trend of error change, and draw the curve of sample mean square error change with the number of hidden layer neurons. Figure 6 .Depend on Figure 6 It can be seen that when the number of neurons in the hidden layer is 11, the mean square error values of the learning and verification samples tend to be stable and almost no longer decrease. Therefore, the number of neurons in the hidden layer is determined to be 11 in the neural network structure, and the overall neural network structure is 6-11-1.
[0095] Compared with the prior art, the present invention uses BP neural network. After training the BP neural network with a large number of samples, the real-time temperature of turbine blade additive can be predicted. The established BP neural network model can reflect the evolution law of blade additive temperature field and can accurately predict the additive temperature under certain processing conditions, thereby providing a theoretical basis for selecting reasonable process parameters and heat input parameters and reducing heat loss. This method is conducive to improving processing quality and reducing production costs, and has certain guiding significance for the actual production process.
[0096] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
[0097] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The schemes in the embodiments of the present application may be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.
[0098] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0099] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0100] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0101] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.
[0102] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.
Claims
1. A method for predicting the real-time temperature field of turbine blade arc additive, characterized in that: include, Establish the temperature field of the turbine blade additive process based on the characteristic dimensions of the profile; Establishing a welding simulation database based on the temperature field, and storing the temperature field calculation results as simulation samples in the database; Build a welding data platform based on welding simulation database and actual welding data; A temperature field prediction model of a BP neural network is established based on the welding data platform; the BP neural network is a multi-layer network for error back propagation, and the error of each previous layer is calculated forward through the error of the output layer; the working process expression of the neurons in the BP neural network is expressed as: Among them, x1~x m Represents m input variables, ω1~ω m is the connection weight, Σ represents the sum, b is the bias value, f is the activation function, and y is the output variable; The input variables of the BP neural network are the blade root form, the blade crown form, the layer thickness, the inlet and outlet steam edge radius, the cross-sectional position and the curve parameter, and the output variable is the temperature; for blades of different models, the characteristic dimensions of the profile, including the inlet and outlet air radius and the chord width, are extracted to determine the input variables, and the basic principle for determining the number of hidden layer nodes is: under the premise of meeting the accuracy, the structure is as compact as possible, that is, the number of hidden layer nodes is as small as possible; after determining the hidden layer, the hyperbolic tangent S-type function is selected as the activation function in the BP neural network learning process, the number of network iterations is set to 100 times, and the network learning rate is set to 0.01; the difference between the learning result and the sample is defined as the loss value, and the number of iterations is determined by the test effect while avoiding overfitting; The training and verification of the established temperature field prediction model include: Data construction: The data obtained from several sets of blade simulation analysis are stored in the database through the server; Data formatting: Reorganize the data by dividing it into two independent data files, one for training and the other for testing, and ensure that they are consistent with the neural network input format; Convergence judgment: When the system converges satisfactorily and the error threshold is met, the ANN parameters are stored for testing. When the system does not converge, the training parameters must be modified until satisfactory results are obtained. Find the optimal solution: Test the solutions with a data test file and use the solution with the smallest error in the prediction process during background processing.
2. A method for predicting the real-time temperature field of turbine blade arc additive according to claim 1, characterized in that: The process of establishing a temperature field for simulating a turbine blade additive process includes establishing a model and defining material properties, dividing units and grids, adding initial and boundary conditions, applying a moving heat source, setting a time step and solving and post-processing results.
3. A method for predicting the real-time temperature field of turbine blade arc additive according to claim 2, characterized in that: The simulated samples are obtained by designing a rotation regression test combination.
4. A method for predicting real-time temperature field of turbine blade arc material addition according to claim 3, characterized in that: The data processing of the sample is to use the mapminmax function in Matlab to normalize the imported training set and test set.
5. A method for predicting the real-time temperature field of turbine blade arc additive according to claim 4, characterized in that: The specific calculation formula of the mapminmax function is expressed as: Where x represents the input value or expected output value in the test sample data; x min Indicates the minimum value in the sample; x max Represents the maximum value in the sample; y represents the normalized value in the range [-1,1], y min =-1;y max =1.
6. A method for predicting the real-time temperature field of turbine blade arc additive according to claim 5, characterized in that: The welding data platform includes a material library, an experimental library, a simulation library and a prediction calculation module.
