Laser cladding molten pool structure parameter prediction method based on dimensionless model

Abstract

The invention belongs to the technical field of laser additive manufacturing, and particularly provides a laser cladding molten pool structure parameter prediction method based on a dimensionless model, which comprises the following steps: S1, determining a process parameter space in a laser cladding process, and establishing a representation function of structural parameters of a molten pool relative to the process parameters; s2, performing dimension analysis and simplification on the representation function, and determining dimensionless parameters; s3, constructing a dimensionless model based on the dimensionless parameters; and S4, on the basis of preset process parameters of the to-be-detected laser cladding molten pool, the structure parameters of the to-be-detected laser cladding molten pool are predicted through the dimensionless model. In the method for predicting the structural parameters of the laser cladding molten pool, a dimensionless model about the laser cladding molten pool is constructed, and a user inputs to-be-measured process parameters to the final dimensionless model based on used process parameters (characteristics of used materials, different powder feeding rates and parameters of laser operation) and the dimensionless model, so as to predict the structural parameters of the laser cladding molten pool. The structure parameters of the molten pool can be predicted.

Description

Technical Field

[0001] The present application belongs to the technical field of additive manufacturing, and in particular relates to a laser cladding molten pool shape prediction method based on a dimensionless model, which is used to efficiently predict the structural parameters of the molten pool and optimize the process parameters. Background Art

[0002] Laser cladding, as a technology for changing the surface properties of materials, has the characteristics of dense cladding structure, high cladding interface bonding strength, easy formation of metallurgical bonding layer, high degree of automation of the cladding process, and can significantly improve the hardness, wear resistance and corrosion resistance of the substrate surface while maintaining the strength and toughness of the substrate. It is widely used in aerospace, automobile manufacturing, petrochemical and mining machinery and other fields. The geometry of the molten pool (width, depth, length) directly affects the forming quality, bonding strength and residual stress distribution of the coating. The defects of laser cladding coatings are complex and diverse, and these defects seriously affect the use of the substrate material. Therefore, how to link process parameters and material properties with the morphology of the laser cladding coating molten pool and apply them to the predictive evaluation of laser cladding coatings is an urgent need in actual engineering production and manufacturing applications. Summary of the Invention

[0003] In order to solve the above technical problems, the present application provides a method for predicting the shape and size of a laser cladding molten pool based on a dimensionless model, comprising:

[0004] S1. Determine the process parameter space during laser cladding and establish a representation function of the structural parameters of the molten pool with respect to the process parameters;

[0005] S2. performing dimension analysis and simplification on the representation function to determine dimensionless parameters;

[0006] S3. constructing a dimensionless model based on the dimensionless parameters;

[0007] S4. Predicting structural parameters of the laser cladding molten pool to be measured using the dimensionless model according to preset process parameters of the laser cladding molten pool to be measured.

[0008] Furthermore, the structural parameters include at least: molten pool width W, molten pool height H;

[0009] The representation function is:

[0010]

[0011] The process parameters in formula (1) are: P is the laser power, v is the scanning speed, R is the laser beam radius, is the powder feeding rate, T * is the melting temperature of the substrate, T0 is the preheating temperature of the substrate, ρ cis the powder density, ρ is the substrate density, C p is the specific heat capacity. S is the structural parameter.

[0012] Furthermore, in step S2, the dimensional analysis and simplification of the representation function includes:

[0013] S211, formula (1) is simplified to:

[0014]

[0015] Among them, E v The energy consumed by heating and melting unit volume of substrate is expressed as:

[0016] E v =ρ·C p ·(T * -T0) Formula (3)

[0017] The dimensionless parameters include:

[0018] Ψ 00 =f(Ψ1,Ψ2) Formula (4)

[0019] Ψ 01 =p(Ψ1,Ψ2) Formula (5)

[0020] in,

[0021]

[0022]

[0023] Ψ 00 Represents the relative area of the molten pool cross section, Ψ 01 Indicates the aspect ratio of the molten pool cross section;

[0024]

[0025] Ψ1 is the relative energy ratio, which represents the laser volume energy density and the energy E absorbed by the substrate when it is heated to the melting point. v Ψ2 is the relative powder feeding mass rate, which means the ratio of the powder feeding mass per unit time to the mass of the powder pre-laid by the laser scanning area with the laser diameter thickness.

[0026] Furthermore, in step S3, a dimensionless model is constructed based on the dimensionless parameters, including:

[0027] S311. Construct a binary linear function model:

[0028] S312. Perform laser cladding experiments based on experimental process parameters of different materials, different powder feeding rates, and different laser operations, and measure and record experimental structural parameters of the molten pool;

[0029] S313. Fit the multiple groups of experimental process parameters and corresponding experimental structural parameters as experimental value sets with the dimensionless parameters of the binary linear function model to determine the value of the linear coefficient of the binary linear function model.

[0030] Furthermore, the binary linear function model includes:

[0031]

[0032] Equation (10) represents the linear relationship between the relative area of the molten pool, the aspect ratio, and the relative laser energy, and Equation (11) represents the linear relationship between the relative powder feeding mass rate; where a1, a2, b1, b2, c1, and c2 are linear coefficients.

[0033] Furthermore, in step S4, predicting the structural parameters of the laser cladding molten pool to be measured includes:

[0034] S411, according to the preset process parameters and the binary linear function model with the determined linear coefficient value, calculate the predicted values of the molten pool width W and the molten pool height H; wherein,

[0035]

[0036] Furthermore, in step S3, a dimensionless model is constructed based on the dimensionless parameters, including:

[0037] S321. Build a dimensionless deep learning model:

[0038] S322. Perform laser cladding experiments based on experimental process parameters of different materials, different powder feeding rates, and different laser operations, and measure and record experimental structural parameters of the molten pool;

[0039] S323. Train a dimensionless deep learning model using an experimental value set, where the experimental value set includes the experimental process parameters and corresponding experimental structure parameters.

[0040] Furthermore, the dimensionless deep learning model is:

[0041]

[0042] Furthermore, predicting the structural parameters of the laser cladding molten pool to be measured includes:

[0043] S421, input the preset process parameters into the trained dimensionless deep learning model, and output The predicted value of

[0044] S422, according to The predicted values of the molten pool width W and the molten pool height H are obtained from the predicted values.

[0045] The above technical solution of the present invention has at least the following beneficial technical effects:

[0046] (1) In the laser cladding molten pool structural parameter prediction method of the present application, a dimensionless model of the laser cladding molten pool is constructed. Based on the process parameters used in the laser cladding process (properties of the material used, different powder feeding rates and parameters of laser operation) and the dimensionless model, the user only needs to input the process parameters to be measured into the final dimensionless model to predict the structural parameters of the molten pool.

[0047] (2) By constructing a binary linear function model, only a few experiments are needed to determine the values of each coefficient of the dimensionless model, saving a lot of time and experimental costs. The dimensionless model can predict the molten pool structure parameters of different materials, different powder feeding rates and different laser operation parameters under the same working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the traditional technology, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0049] Figure 1 This is a flow chart of a method for predicting laser cladding molten pool structural parameters based on a dimensionless model in one embodiment of the present application.

[0050] Figure 2 is the experimental value of WH / R^2 and the dimensionless parameter Ψ in Example 1 of this application 00 The fitting graph of .

[0051] Figure 3 is the experimental value of H / W and the dimensionless parameter Ψ in Example 1 of this application 01 The fitting graph of .

[0052] Figure 4 is the dimensionless parameter Ψ obtained by deep learning in Example 2 of this application 00 Scatter plot of experimental predictions versus experimental values.

[0053] Figure 5 is the dimensionless parameter Ψ obtained by deep learning in Example 2 of this application 01 Scatter plot of experimental predictions versus experimental values. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, each embodiment of the present application will be described in detail below with reference to the accompanying drawings. However, it will be understood by those skilled in the art that in each embodiment of the present application, many technical details are proposed to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in the present application can be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation on the specific implementation of the present application. The various embodiments can be combined with each other and referenced to each other under the premise of no contradiction.

[0055] Definition of noun:

[0056] Laser cladding technology, as an advanced surface modification and repair method, has been widely used in many industries. For example, by cladding a layer of higher performance material on the surface of a component, it can extend the service life of the component and reduce replacement costs.

[0057] The laser cladding process involves locally melting metal powder or wire on the substrate surface, metallurgical bonding, and rapid solidification. The process can be roughly divided into the following steps:

[0058] Pre-laying and powder feeding: Pre-laying metal powder / wire on the substrate surface or feeding it through a feeding system to form a predetermined material layer.

[0059] Laser irradiation and melting: After being focused, the laser beam is irradiated onto the surface of the preset material and the substrate. The high-energy laser instantly melts them locally to form a molten pool. At the same time, the partial melting of the substrate facilitates metallurgical bonding.

[0060] Mixing and metallurgical reaction: In the molten pool, metal powder is mixed with part of the base material, and a new alloy layer is formed through metallurgical reaction. Its composition and structure can be designed and optimized by humans and computer systems with specific calculation models to meet specific performance requirements.

[0061] Rapid solidification: As the laser beam moves, the molten pool cools and solidifies rapidly, forming a dense cladding layer and forming a strong metallurgical bond with the substrate.

[0062] At present, the existing technology has the characteristics of high nonlinearity, multi-scale, multi-parameter and multiple physical field coupling. The prediction of molten pool structural parameters mainly relies on numerical simulation or a large amount of experimental data, which consumes a lot of computing resources. The process parameters need to be repeatedly tried and errored, which is costly and inefficient.

[0063] To solve the above problems, an embodiment of the present application provides a method for predicting the structural parameters of a laser cladding molten pool based on a dimensionless model, which specifically includes:

[0064] S1. Determine a process parameter space based on various process parameters frequently used in the laser cladding process, and establish a representation function of the structural parameters of the molten pool with respect to the process parameters based on the process parameter space; wherein the process parameters include the characteristics of the material used, the powder feed rate of the metal powder, and the parameters of the laser operation.

[0065] S2. Simplify the representation function and perform dimension analysis to determine the dimensionless parameters. After determining the dimensionless parameters, the dimensionless parameters can be further optimized.

[0066] S3. Construct a dimensionless model based on dimensionless parameters.

[0067] S4. Based on the preset process parameters of the laser cladding molten pool to be measured, the structural parameters of the laser cladding molten pool to be measured are predicted using a dimensionless model; wherein the structural parameters include at least: a molten pool width W and a molten pool height H.

[0068] In the laser cladding molten pool structural parameter prediction method of the present application, a dimensionless model of the laser cladding molten pool is constructed, and based on the process parameters used in the laser cladding process (the characteristics of the material used, different powder feeding rates and laser operation parameters) and the experimental value set of the molten pool structural parameters are obtained, the values of the dimensionless model coefficients are verified and determined; the final dimensionless model is different from the dimensionless models of the prior art that use large amounts of data or neural networks. The user only needs to input the process parameters to be measured into the final dimensionless model, and the predicted molten pool structural parameters can be obtained through simple and direct calculations.

[0069] In a specific embodiment of the present application, the material properties of the same substrate are set to be unchanged during laser operation; the number of initial process parameters is set to 9, which are the main factors affecting the structure and shape of the molten pool. The function representing the structural parameters and process parameters can be expressed as:

[0070]

[0071] The process parameters in formula (1) are: P is the laser power; v is the scanning speed; R is the laser beam radius; is the powder feeding rate; T * is the melting temperature of the substrate; T0 is the preheating temperature of the substrate; ρ c is the powder density; ρ is the substrate density; C p is the specific heat capacity. S represents the structural parameters, which include the molten pool width W and the molten pool height H.

[0072] The units of the 9 process parameters in the function are shown in Table 1 below:

[0073] Table 1: Symbols of process parameters, their corresponding physical meanings and units

[0074]

[0075] The preheating temperature in Table 1, related to substrate preheating, refers to the heating of the substrate by the laser. Its core function is to reduce thermal gradients and alleviate thermal stresses. Its advantages include: reducing crack tendency. Preheating can reduce the temperature gradient between the cladding layer and the substrate, reducing thermal stress caused by differences in thermal expansion coefficients, and thus significantly reducing crack sensitivity. It also improves the wettability and bonding quality of the molten pool. The preheated substrate surface can enhance the wettability of the molten powder and promote the metallurgical bonding between the cladding layer and the substrate. It also optimizes the microstructure and properties. Preheating can slow the solidification rate, refine the grain size of the cladding layer, and evenly distribute it, thereby improving the wear resistance and corrosion resistance of the coating. It also reduces the dilution rate. Preheating reduces the reflection of laser energy from the substrate surface and improves energy absorption efficiency, thereby reducing the substrate melting depth at the same power and controlling the dilution rate.

[0076] Furthermore, in step S2, the representation function is dimensionally analyzed and simplified, including:

[0077] S211. Simplify the formula (1) corresponding to the expression function to further reduce the number of factors in the expression function, specifically:

[0078] Formula (1) is simplified to:

[0079]

[0080] Among them, E v The energy consumed by heating and melting unit volume of substrate is expressed as:

[0081] E v =ρ·C p ·(T * -T0) formula (3).

[0082] In one embodiment of the present application, in step S2, performing dimensional analysis and simplification on the representation function further includes:

[0083] S22. Perform dimensional analysis and non-dimensionalize equation (2) to obtain the following dimensionless parameters:

[0084] Ψ 00 =f(Ψ1,Ψ2) Formula (4)

[0085] Ψ 01 =p(Ψ1,Ψ2) Formula (5)

[0086] in,

[0087]

[0088] In formula (6) and formula (7), Ψ 00Represents the relative area of the molten pool cross section, Ψ 01 Represents the aspect ratio of the melt pool cross section.

[0089]

[0090] Ψ1 represents the laser volume energy density and the energy E absorbed by the substrate when it is heated to the melting point v Ψ2 represents the ratio of the powder delivery mass per unit time to the mass of the powder pre-laid by the laser scanning area with the laser diameter thickness, that is, the relative powder delivery mass rate.

[0091] In the specific embodiment of the present application, based on the aforementioned dimensionless parameter Ψ 00 ,Ψ 01 , Ψ1, Ψ2, two types of dimensionless models were constructed, one of which is a binary linear function model and the other is a dimensionless deep learning model, both of which can predict the structural parameters of the molten pool. The following describes the construction process of the two types of dimensionless models and the process of predicting the structural parameters of the molten pool, as shown in Examples 1 and 2.

[0092] Example 1

[0093] In step S3, a dimensionless model is constructed based on the dimensionless parameters, including:

[0094] S311. Construct a binary linear function model:

[0095]

[0096] Equation (10) expresses the relative area and aspect ratio of the melt pool and the relative laser energy. Equation (11) expresses the linear relationship between the relative powder feeding mass rate, which generally obeys the laws of conservation of energy and mass. Where a1, a2, b1, b2, c1, and c2 are linear coefficients.

[0097] S312. Perform laser cladding experiments based on experimental process parameters of different materials, different powder feeding rates, and different laser operations, and measure and record experimental structural parameters of the molten pool;

[0098] S313. Fitting the multiple groups of experimental process parameters and corresponding experimental structural parameters as experimental value sets with the dimensionless parameters of the binary linear function model to determine the values of the linear coefficients of the binary linear function model.

[0099] In one embodiment of the present application, in step S3, the process of performing the laser cladding experiment is consistent with the laser process described in the aforementioned definition of terms. Experiments are performed and data is recorded based on different materials, different powder feeding rates, and process parameters of laser operation. The more experiments, the better. The experimental value set formed is as shown in Table 2 below:

[0100] Table 2 Experimental value set obtained from laser cladding experiment

[0101]

[0102] Table 2 records the experimental process parameters for each substrate, including Hastelloy alloy and 316 stainless steel, with NI60 metal powder used for both. The experimental process parameters for each substrate at different powder feed rates and laser operating parameters are also recorded, as well as the experimental structural parameters (melt pool width and depth) obtained through measurement. The experimental value set in Table 2 is fitted with the dimensionless parameters in the binary linear function model using the following fitting method:

[0103] (1) The experimental value set in Table 2 and formula (10) were imported into the origin software for analysis. After manually selecting the multiple linear regression option, the software analyzed the data and obtained the values of the linear coefficients a1, b1, and c1 in the binary linear function model. The fitting results were: a1 = 3.89778, b1 = 1.13582, c1 = 2.60237.

[0104] (2) The experimental value set in Table 2 and formula (11) were imported into the origin software for analysis. After manually selecting the multiple linear regression option, the software analyzed the data and obtained the values of the linear coefficients a2, b2, and c2 in the binary linear function model. The fitting results were: a2 = 0.02599, b2 = 0.07845, c2 = 0.08868.

[0105] The purpose of the above fitting is that since every three groups of experimental values can determine the value of a group of linear coefficients, the values of multiple groups of linear coefficients are very similar, and a group of relatively average and highly applicable linear coefficients can be obtained through fitting.

[0106] The binary linear function model with the determined linear coefficient value is obtained in the above manner. In actual production, the preset process parameters are substituted into the model, and the structural parameters are predicted by direct calculation. In order to verify the accuracy of the binary linear function model, in this embodiment, a part of the experimental value set is used as verification data, and the experimental values of WH / R^2 and H / W are calculated one by one; and this part of the verification data is input into the binary linear function model after the linear coefficient value is determined, and the data obtained by the WH / R^2 and H / W fitting regression are output. The results are as follows Figure 2 and Figure 3 As shown, the experimental values of WH / R^2 and H / W tend to be consistent with the data obtained by the fitting regression of WH / R^2 and H / W, indicating that the actual experimental data are relatively close to the structural parameters predicted by the binary linear function model in this embodiment.

[0107] In step S4, the structural parameters of the laser cladding molten pool to be measured are predicted, including:

[0108] S411. When the user provides preset parameters, the molten pool width W and molten pool height H are calculated based on the preset process parameters and the binary linear function model with the determined linear coefficient value, as follows:

[0109] The predicted values of the molten pool width W and molten pool height H can be calculated by equations (10) and (11);

[0110]

[0111] The binary linear function model in this embodiment only requires several experiments to determine the value of the linear coefficient of the dimensionless model, saving a lot of time and experimental costs. The dimensionless model can predict the structural parameters of the molten pool of different materials, different powder feeding rates and different laser operating parameters under the same working conditions.

[0112] Example 2

[0113] In this embodiment, in step S3, the dimensionless model is a dimensionless deep learning model, and the construction steps include:

[0114] S321. Build a dimensionless deep learning model:

[0115]

[0116] S322: Conduct laser cladding experiments based on experimental process parameters for different materials, different powder feed rates, and different laser operations, and measure and record experimental structural parameters of the molten pool. The experimental structural parameters in this embodiment can be obtained in the same manner as in step S312 of Example 1, or directly using the experimental value set.

[0117] S323. Train a dimensionless deep learning model using experimental value sets consisting of multiple sets of experimental process parameters and corresponding experimental structural parameters. Preferably, the experimental value set is divided into a training set and a validation set, with the training set accounting for 80% and the validation set accounting for 20%. The input is the process parameters and structural parameters and their values as shown in Table 2, and the output is the predicted values of WH / R^2 and H / W. The predicted values of WH / R^2 and H / W are further used to calculate the predicted values of the melt pool depth H and the melt pool width W.

[0118] During the model training process of this embodiment: for each set of experimental values, the model will also output a corresponding set of experimental prediction values of WH / R^2 and H / W; at the same time, the model will also calculate the experimental values of WH / R^2 and H / W corresponding to each set of experimental values one by one (equivalent to manually calculating the experimental values of WH / R^2 and H / W based on the process parameters and structural parameters in Table 2). Finally, as Figure 4 and Figure 5As shown in the scatter plot, the model compares and visualizes the experimental predicted values of WH / R^2 and H / W with the experimental values of WH / R^2 and H / W; repeated training makes the two converge.

[0119] In this embodiment, the dimensionless deep learning model construction and training further involves the following steps in machine implementation:

[0120] 1. Loading and preprocessing of experimental set data:

[0121] Function: load_data

[0122] Use pandas to load the Excel file of Table 2.

[0123] Define the input column (m(g / min)P(W)R(mm)υ(mm / s)W(mm)D(mm)H(mm)ρρ c Cp T*T 0 ) and output columns (WH / R^2, H / W) and check whether these columns exist in the data. Ensure that the data is formatted correctly to avoid program crashes due to incorrect column names. Define the input and output columns in advance to facilitate subsequent operations.

[0124] If the column names do not match, an exception is thrown.

[0125] Add some new features generated by feature engineering (such as etc.) to enhance the performance of the model.

[0126] 2. Small data enhancement:

[0127] Function: augment_data

[0128] The original experimental value set is augmented to generate new samples by adding Gaussian noise to each data.

[0129] Control the noise range to ensure that the amplified data still conforms to physical constraints.

[0130] Deep learning models typically require large amounts of data to achieve good generalization. When the amount of data is limited, data augmentation techniques can be used to expand the training set. This approach adds noise while limiting the range to ensure the rationality of the augmented data.

[0131] 3. Data standardization:

[0132] Function: standardize_data

[0133] StandardScaler is used to standardize the input features and target variables (mean is 0, standard deviation is 1); the input features are (m(g / min)P(W)R(mm)υ(mm / s)W(mm)

[0134] D(mm)H(mm)ρρ c Cp T*T 0 ); the target variable is (WH / R^2,H / W).

[0135] Standardizing the experimental data can accelerate the convergence of the neural network and prevent numerical differences between features of different dimensions from affecting model performance. The target variable also needs to be standardized to ensure a more stable loss function calculation.

[0136] 4. Build a deep learning model

[0137] Function: build_model

[0138] Use tensorflow.keras to build a simple fully connected neural network:

[0139] Input layer: set according to the input feature dimension.

[0140] Hidden layer: two layers, with 64 and 32 neurons respectively, and the activation function is ReLU.

[0141] Output layer: set according to the target variable dimension, no activation function (regression task).

[0142] Compile the model, use Adam as the optimizer, and mean_squared_error as the loss function.

[0143] Fully connected neural network: suitable for solving nonlinear regression problems and can capture the complex relationship between input features and target variables.

[0144] ReLU activation function: introduces nonlinearity to improve the model's expressiveness.

[0145] Adam optimizer: Adaptive learning rate, suitable for most deep learning tasks.

[0146] MSE loss function: measures the square error between the predicted value and the true experimental value, suitable for regression tasks.

[0147] 5. Model training and validation:

[0148] Function: train_model

[0149] After standardizing the data, divide it into training and test sets.

[0150] Call build_model to build the model.

[0151] The model is trained for 100 epochs with a batch size of 1 while reserving 20% of the data for validation.

[0152] Evaluate model performance and calculate R 2 and MSE.

[0153] Training set / test set split: ensure the model's generalization ability on unseen data.

[0154] Validation set: monitors overfitting during training.

[0155] R 2 and MSE: Commonly used regression model evaluation indicators, measuring the explanatory power and prediction error of the model respectively.

[0156] 6. Results visualization

[0157] Function: plot_predictions

[0158] Draw a scatter plot of the true experimental value and the experimental predicted value for each target variable (WH / R^2 and H / W) and add a regression line.

[0159] Display R 2 Use the value as part of the title. Use a scatter plot + regression line to visually display the model prediction effect.

[0160] R 2 Value: A measure of the goodness of fit of the model.

[0161] Detailed explanation of the dimensionless deep learning model

[0162] Model structure

[0163] First layer (input layer + hidden layer): The input dimension is determined by input_dim (i.e. the number of input features). Set 64 neurons and the activation function is ReLU.

[0164] The second layer (hidden layer): 32 neurons are set, and the activation function is ReLU.

[0165] The third layer (output layer): The output dimension is determined by output_dim (i.e. the number of target variables). There is no activation function and the predicted value is directly output.

[0166] Compile the model: Use Adam as the optimizer and an adaptive learning rate algorithm, suitable for most deep learning tasks. Initially set the learning rate to 0.001. Use the Mean Squared Error (MSE) loss function, a common loss function for regression tasks that measures the squared error between the predicted value and the actual experimental value.

[0167] Training the model:

[0168] epochs: training 100 times.

[0169] Batch_size: Only one piece of data is used for each training (small batch gradient descent).

[0170] validation_split: Reserve 20% of the data for validation.

[0171] verbose: silent mode, does not display the training process.

[0172] The prediction method proposed in this example involves multiple process parameters, which are simplified into a number of dimensionless parameters after dimensional analysis, focusing on key parameters. Before actual production, a dimensionless deep learning model can predict the structural parameters of the laser cladding coating melt pool formed by these process parameters in advance, saving costs, shortening cycle time and data analysis, accelerating process optimization, and providing guidance for process and product optimization.

[0173] The specific features, structures, materials, or characteristics described in the present invention may be combined in any suitable manner in any one or more embodiments or examples.

[0174] It should be understood that the above-mentioned specific embodiments of the present application are merely illustrative or explain the principles of the present application and do not constitute a limitation of the present application. Therefore, any modifications, equivalent substitutions, improvements, etc. made without departing from the spirit and scope of the present application should be included in the scope of protection of the present application. In addition, the claims attached hereto are intended to cover all variations and modifications that fall within the scope and boundaries of the appended claims, or the equivalent forms of such scope and boundaries.

Claims

1. A method for predicting the structural parameters of a laser cladding molten pool based on a dimensionless model, characterized in that: include: S1. Determine the process parameter space during laser cladding and establish a representation function of the structural parameters of the molten pool with respect to the process parameters; S2. performing dimension analysis and simplification on the representation function to determine dimensionless parameters; S3. constructing a dimensionless model based on the dimensionless parameters; S4. Predicting structural parameters of the laser cladding molten pool to be measured using the dimensionless model according to preset process parameters of the laser cladding molten pool to be measured.

2. The prediction method according to claim 1, characterized in that The structural parameters include at least: molten pool width W, molten pool height H; The representation function is: The process parameters in formula (1) are: P is the laser power, v is the scanning speed, R is the laser beam radius, is the powder feeding rate, T * is the melting temperature of the substrate, T0 is the preheating temperature of the substrate, ρ c is the powder density, ρ is the substrate density, C p is the specific heat capacity; S is the structural parameter.

3. The prediction method according to claim 2, characterized in that In step S2, the dimensional analysis and simplification of the representation function includes: S211, formula (1) is simplified to: Among them, E v The energy consumed by heating and melting unit volume of substrate is expressed as: E v =ρ·C p ·(T * -T0) formula (3); The dimensionless parameters include: Ψ 00 = f(Ψ1, Ψ2), Equation (4) Ψ 01 = p(Ψ1, Ψ2) Equation (5) in, Ψ 00 Represents the relative area of the molten pool cross section, Ψ 01 Indicates the aspect ratio of the molten pool cross section; Ψ1 is the relative energy ratio, which represents the laser volume energy density and the energy E absorbed by the substrate when it is heated to the melting point. v Ψ2 is the relative powder feeding mass rate, which means the ratio of the powder feeding mass per unit time to the mass of the powder pre-laid by the laser scanning area with the laser diameter thickness.

4. The prediction method according to claim 3, characterized in that In step S3, a dimensionless model is constructed based on the dimensionless parameters, including: S311. Construct a binary linear function model: S312. Perform laser cladding experiments based on experimental process parameters of different materials, different powder feeding rates, and different laser operations, and measure and record experimental structural parameters of the molten pool; S313. Using multiple groups of experimental process parameters and corresponding experimental structural parameters as experimental value sets, fitting them with the dimensionless parameters of the binary linear function model, and determining the values of the linear coefficients of the binary linear function model.

5. The prediction method according to claim 4, characterized in that The binary linear function model includes: Equation (10) represents the linear relationship between the relative area of the molten pool, the aspect ratio, and the relative laser energy, and Equation (11) represents the linear relationship between the relative powder feeding mass rate; where a1, a2, b1, b2, c1, and c2 are linear coefficients.

6. The prediction method according to claim 5, characterized in that In step S4, predicting the structural parameters of the laser cladding molten pool to be measured includes: S411, according to the preset process parameters and the binary linear function model with the determined linear coefficient value, calculate the predicted values of the molten pool width W and the molten pool height H; wherein, 7. The prediction method according to claim 4, characterized in that In step S3, a dimensionless model is constructed based on the dimensionless parameters, including: S321. Build a dimensionless deep learning model: S322. Perform laser cladding experiments based on experimental process parameters of different materials, different powder feeding rates, and different laser operations, and measure and record experimental structural parameters of the molten pool; S323. Train a dimensionless deep learning model using an experimental value set, where the experimental value set includes the experimental process parameters and corresponding experimental structure parameters.

8. The prediction method according to claim 7, characterized in that The dimensionless deep learning model is:

9. The prediction method according to claim 8, characterized in that Predicting the structural parameters of the laser cladding molten pool to be measured includes: S421, input the preset process parameters into the trained dimensionless deep learning model, and output The predicted value of S422, according to The predicted values of the molten pool width W and the molten pool height H are obtained from the predicted values.

Images