Artificial Intelligence-Based Method for Laser Directed Energy Deposition of Nickel-Based Single Crystals

Abstract

The present invention discloses a method for laser directed cladding of nickel-based single crystals based on artificial intelligence, which controls the growth of nickel-based single crystals along the correct epitaxial growth direction, i.e., the [001] direction of nickel-based single crystals. Specifically, it includes determining data, searching for data and forming a training data set, determining a threshold and screening the training data set according to the threshold, processing the data through a BP neural network to obtain the optimal laser cladding angle, and determining the laser cladding trajectory of nickel-based single crystal superalloys in the above manner, presetting the optimal laser cladding trajectory, and performing laser cladding on the pre-treated substrate material at the optimal laser cladding angle. The method of the present invention abandons the complex method of using MATLAB calculations, can independently determine the laser cladding angle and the epitaxial growth direction, makes it faster and more convenient to determine the dendrite growth direction, cladding angle and cladding trajectory, has a larger single crystal area for the repaired nickel-based single crystal superalloy, the single crystal orientation is closer to the substrate, and the cracking probability is lower.

Description

Technical Field

[0001] The present invention belongs to the technical field of nickel-based single crystal additive manufacturing and repair, and relates to a method for laser directed cladding of nickel-based single crystals using artificial intelligence to repair nickel-based single crystal superalloys. Background Art

[0002] The nickel-based single crystal superalloy repair technology can repair the worn parts and cracks at the tip of the turbine blade, greatly extending the service life of the blade. Moreover, compared with replacing a new blade, using the repair technology can greatly reduce the maintenance cost of the engine.

[0003] Due to the single crystal property of the nickel-based single crystal superalloy, it is required that the dendrite growth direction during repair is as consistent or similar as possible with the substrate direction, so as to have good connection ability with the substrate to be repaired and maintain good performance. That is, an ideal single crystal blade repair technology should be able to produce a single crystal or columnar crystal structure similar to the single crystal blade tissue.

[0004] The laser cladding repair technology has become the preferred technology for nickel-based single crystal superalloy repair because of its small heat affected zone, good mechanical properties of the formed parts, and the ability to achieve automatic control.

[0005] For a long time, the dendrite orientation of repairing nickel-based single crystal superalloys by laser cladding has been a research hotspot attracting much attention. Studying the epitaxial growth direction and the formation of the microstructure of nickel-based single crystal by mathematical models and experimental methods has become the main research content.

[0006] However, the existing theoretical models are cumbersome and complex, and not systematic. Even when using software for simulation and calculation, it is very laborious, often requiring a long time of human effort, and it is also difficult to select the optimal orientation. There is no good method for determining the cladding angle and cladding trajectory of laser cladding.

[0007] With the continuous expansion and improvement of the mathematical model, the direction of dendrite epitaxial growth has become clearer. For example, MATLAB and simulation software can be used for calculation and simulation, but the huge amount of calculation data and calculation process are still a relatively large problem.

[0008] In recent years, artificial intelligence has gradually expanded. Artificial intelligence can be applied to determine the dendrite orientation of epitaxial growth of nickel-based single crystal superalloys, and the cladding angle and cladding trajectory are determined through continuous training and neural network screening. Summary of the Invention

[0009] The object of the present invention is to provide a method for laser directed cladding of nickel-based single crystals based on artificial intelligence. The method provided by the present invention abandons the complex method of using MATLAB calculations, and can independently determine the laser cladding angle and the epitaxial growth direction, making it faster and more convenient to determine the dendrite growth direction, cladding angle and cladding trajectory. The single crystal area of the repaired nickel-based single crystal superalloy is larger, the single crystal orientation is closer to the substrate, and the cracking probability is lower.

[0010] The method for laser directed cladding of nickel-based single crystals based on artificial intelligence according to the present invention is to control the nickel-based single crystal to grow along the correct epitaxial growth direction by artificial intelligence. The correct epitaxial growth direction specifically refers to the

[001] direction of the nickel-based single crystal, and specifically includes:

[0011] 1. Using a Cartesian coordinate system, setting the positive direction of the y-axis as the laser scanning direction and the z-axis direction as the

[001] direction, and determining the laser cladding angle of the nickel-based single crystal superalloy based on artificial intelligence

[0012] 1). Determine the data, search for the data and form a training data set

[0013]

[0014]

[0015]

[0016] Among them:

[0017] The melting temperature T, thermal conductivity k, and thermal diffusivity α of the material properties are constants;

[0018] P is the laser power absorbed by the substrate, W;

[0019] V b is the laser scanning speed, mm / s;

[0020] T0 is the substrate temperature, K;

[0021] The size of the molten pool is determined by x, y, and z, mm.

[0022] Input the melting temperature T, thermal conductivity k, thermal diffusivity α of the material properties, the laser power P absorbed by the substrate, and the laser scanning speed V b , the substrate temperature T0, then the size of the molten pool, temperature gradient G n and its components G x , G y , G z .

[0023] For different materials and different processes, different molten pools and temperature gradient characteristics will be obtained.

[0024] Then determine G according to the following formula n / V, where for nickel-based single crystals, the value of the material property n is 3.4:

[0025]

[0026] in:

[0027] hkl refers to the crystal plane index of nickel-based single crystal,

[0028] G n is the temperature gradient;

[0029] V b is the laser scanning speed;

[0030] G x is the component of the temperature gradient on the X-axis.

[0031] On the right side of the equation, only the deviation angle of the preferred orientation dendrite is unknown. n , V b , G x Input, ψ hkl As an unknown parameter, it is uniformly valued and needs to be valued in a specific preferred direction, that is, <001> The six dendrite growth directions of the crystal plane family are output as G in six directions. n / V collection.

[0032] Intelligently select G in one of the six directions n / V, the judgment basis is that the absolute value of G is required to be the largest, then at this time it is determined to be the optimal growth direction.

[0033] Automatically obtain an experimental data set of laser cladding of the nickel-based single crystal high-temperature alloy through big data;

[0034] The laser cladding angle μ and the dendrite growth deviation angle ψ are constructed using the experimental data set. hkl A mapping relationship data set, using the mapping relationship data set as a first training data set;

[0035] A mapping relationship data set between the laser cladding angle μ and the single crystal area percentage is constructed through the experimental data set, and the mapping relationship data set is used as the second training data set.

[0036] 2) Determine the threshold and filter the training data set according to the threshold

[0037] K CET As a first reference value, G n / V and K CET By comparison, we can get G that can be grown into a single crystal. n / V collection.

[0038]

[0039] Determine the deviation threshold of the dendrite growth deviation angle according to the performance parameters, and each G in the set n / V corresponds to a dendrite growth deviation angle ψ hkl .

[0040] The performance parameters selected are to determine K CET Parameter value of a(m -1 s -1 ) and n are material related parameters, N0 is the nucleation density in the liquid (m -3 ), parameter φ c It is defined as the critical value of the area fraction in front of the premature solidification front consisting of newly nucleated grains.

[0041] The deviation threshold is the above G n / V The collection range corresponding to the collection.

[0042] The first training data set and the second training data set are screened based on the above deviation threshold to obtain a training data set for laser cladding, while data outside the deviation threshold is eliminated.

[0043] 3) Process the data through BP neural network to obtain the optimal laser cladding angle

[0044] The calculated dendrite growth deviation angle of the optimal growth direction and the training data set are input into the neural network model, which performs optimization and outputs the preset laser cladding angle μ and the angle ψ between the molten pool boundary normal phase and the specific preferred crystal orientation. hkl .

[0045] 2. According to step 1, artificial intelligence determines the laser cladding trajectory of nickel-based single crystal high-temperature alloy

[0046] 1) Determine data, search data and form a training data set

[0047] Automatically obtain an experimental data set of laser cladding of the nickel-based single crystal high-temperature alloy through big data;

[0048] A mapping relationship dataset between the molten pool size and the cladding trajectory is constructed using the experimental dataset, and the mapping relationship dataset is used as the third training dataset.

[0049] 2) Determine the threshold and filter the training data set according to the threshold

[0050] The deviation threshold of the dendrite growth deviation angle is determined according to the performance parameters, and the selected performance parameters are the performance parameters for determining the size of the molten pool:

[0051]

[0052] in:

[0053] The material properties melting temperature T, thermal conductivity k, and thermal diffusivity α are constants;

[0054] P is the laser power absorbed by the substrate, W;

[0055] V b is the laser scanning speed, mm / s;

[0056] T0 is the substrate temperature, K;

[0057] The size of the molten pool is determined by x, y, and z, mm.

[0058] Input material properties melting temperature T, thermal conductivity k, thermal diffusion coefficient α, laser power P absorbed by the substrate, laser scanning speed V b , substrate temperature T0, the output x value range is x min ~x max , the value range of y is y min ~y max , then the deviation threshold is a set within the range of values of x and y.

[0059] The third training data set is screened based on the deviation threshold to obtain a training data set for the laser cladding trajectory, while data outside the deviation threshold is eliminated.

[0060] 3) Process the data through BP neural network to obtain the optimal laser cladding trajectory

[0061] The calculated melt pool size, the preset laser cladding angle μ, the angle ψ between the melt pool boundary normal and the specific preferred crystal orientation hkl The shape and size of the substrate to be repaired and the training data set are input into the BP neural network model, and the neural network model outputs the optimal cladding trajectory.

[0062] 3. Preset the optimal laser cladding trajectory and perform laser cladding on the pretreated base material according to the optimal laser cladding angle

[0063] A laser cladding angle is preset, the angle between the laser incident direction and the nickel-based single crystal high-temperature alloy substrate is adjusted, and laser cladding is performed on the nickel-based single crystal high-temperature alloy according to the preset laser cladding trajectory.

[0064] Wherein, the base material is pre-treated before laser cladding, including grinding, polishing and cleaning the base surface.

[0065] Specifically, the neural network model, namely the neural network model in machine learning, is a complex neural network system formed by a large number of simple processing units (called neurons) widely interconnected. It reflects many basic characteristics of the human brain function and is a highly complex non-linear dynamic learning system. The neural network model is described based on the mathematical model of neurons. The neural network model obtains highly abstract output information through self-learning. Among them, the BP neural network is a neural network learning algorithm. The network learns in the way of teacher teaching, constantly correcting the data, and this process is repeated alternately until the error tends to the given minimum value, then the learning process is completed.

[0066] The present invention uses the screened data set for laser cladding and the data set for laser cladding angle as the training data of the BP neural network model to train the model. The neural network model obtains experience through data training and continuous self-learning. The dendrite deviation angle in the optimal growth direction, the size of the molten pool, and the size of the contour of the substrate material to be repaired are used as the input information of the neural network model. Through continuous learning, the optimal laser cladding angle corresponding to the dendrite deviation angle in the optimal growth direction, the size of the molten pool, and the optimal cladding trajectory corresponding to the size of the contour of the shape to be repaired are obtained, as the optimal angle and the optimal cladding trajectory that meet the alloy cladding requirements.

[0067] Compared with the prior art, the method constructed by the present invention can use artificial intelligence to determine the cladding angle to achieve the control of the directional growth of the tissue, and use artificial intelligence to obtain the cladding trajectory, which can greatly improve the cladding quality and is applicable to the process technology of nickel-based single crystal additive manufacturing and repair. Brief Description of the Drawings

[0068] Figure 1 It is a diagram of the directionally grown crystal structure after laser cladding using the method of the present invention in Example 1.

[0069] Figure 2 It is a diagram of the directionally grown crystal structure after laser cladding without using the method of the present invention in Comparative Example 1. Embodiments

[0070] The following further describes in detail the specific embodiments of the present invention in conjunction with the drawings and embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention, so that those skilled in the art can better understand and utilize the present invention, rather than limiting the protection scope of the present invention.

[0071] In the embodiments and comparative examples of the present invention, the production processes, experimental methods or detection methods involved, unless otherwise specified, are all conventional methods in the prior art, and their names and / or abbreviations are all conventional names in the art, which are very clear and definite in the relevant application fields. Those skilled in the art can understand the conventional process steps according to the names and apply the corresponding equipment, and implement them under conventional conditions or conditions recommended by the manufacturer.

[0072] There are no special restrictions on the sources of various instruments, equipment, raw materials or reagents used in the embodiments of the present invention. They are all conventional products that can be obtained through regular commercial channels, and can also be prepared according to the conventional methods well-known to those skilled in the art.

[0073] During the process of laser directed energy deposition of nickel-based single crystals, during the rapid solidification of the molten pool, columnar dendrites will grow along six specific dendrite growth directions represented as

[100] , [-100],

[010] , [0-10],

[001] and [00-1]. However, due to the substrate orientation, there is a certain angle between the theoretical dendrite growth direction and the actual growth direction.

[0074] In order to enable the nickel-based single crystal to successfully epitaxially grow along the

[001] direction, which is one of the six directions more consistent with the substrate growth direction, a theoretical model is established to determine the actual optimal growth direction.

[0075] Set the direction of laser cladding as the y-axis direction and the laser incident direction as the z-axis direction, which is also the

[001] direction of one of the six directions, and represent the crystal plane indices of the nickel-based single crystal with hkl. After fixing the substrate orientation and the cladding process, study the influence of the cladding angle on dendrite growth.

[0076] First of all, the major premise is that the substrate to be epitaxially grown must grow into a single crystal and must follow the following theoretical basis:

[0077]

[0078] The CET transformation is the columnar crystal - equiaxed crystal transformation. When the CET transformation occurs, that is, G n / V < K CET , the columnar crystals turn into equiaxed crystals, forming a non-directional structure, that is, a non-single crystal. Therefore, in actual operation, it is not desired to have the CET transformation, but it is desired that the calculated G n / V > K CET . [[ID=3l]]

[0079] Among them, G is the temperature gradient at the solidification front, V is the solidification rate at the front of the columnar dendrites; K CETis the critical value of the CET and is a material - related constant. a and n are material parameters, taking the value of 3.4 for nickel - based single - crystal superalloys; N0 is the nucleation density in the liquid; φ c is the critical value of the area fraction of the front of the premature solidification front composed of newly nucleated grains.

[0080] These parameters include K CET which is a constant when the clad substrate material is determined, while G n / V, that is, the ratio of the temperature gradient to the solidification rate of the solidification front to be calculated, is used to judge whether the dendrites grow epitaxially into single crystals through calculation.

[0081] First of all, the present invention needs to determine the laser cladding angle of nickel - based single - crystal superalloys based on artificial intelligence.

[0082] 1), Determine data, search for data and form a training data set

[0083] The mathematical model is described as follows according to the temperature - field calculation expression:

[0084]

[0085]

[0086]

[0087] Where:

[0088] The material properties, melting temperature T, thermal conductivity k, thermal diffusivity α are constants;

[0089] P is the laser power absorbed by the substrate, W;

[0090] V b is the laser scanning speed, mm / s;

[0091] T0 is the substrate temperature, K;

[0092] The size of the molten pool is determined by x, y, z, mm.

[0093] Input the material properties, melting temperature T, thermal conductivity k, thermal diffusivity α, the laser power P absorbed by the substrate, the laser scanning speed V b , and the substrate temperature T0, then the size of the molten pool can be calculated and modeled, the grid can be divided, and the temperature gradient and solidification rate of each point can be obtained.

[0094] When the process parameters (laser power, scanning speed, powder feeding speed, melting temperature, thermal conductivity, diffusion coefficient, etc.) remain unchanged, the shape of the molten pool and the related temperature gradient also remain unchanged. Therefore, based on the above formula, the size of the molten pool, the normal temperature gradient G of the solidification front nand its component G x , G y , G z , and the normal solidification rate V of the solidification front n . Among them, G x , G y and G z are the temperature gradients in the x, y, and z-axis directions respectively.

[0095] The normal solidification rate V in front of the solidification front n is geometrically connected to the laser moving speed V through the angle θ b ; the normal temperature gradient G n is consistent with the normal solidification rate V n , while the laser scanning speed V b is parallel to G x .

[0096]

[0097]

[0098] The direction obtained at this time is not the actual growth direction of the dendrite, but the direction obtained by theoretical calculation. There is a certain angle ψ between its direction and the actual growth direction of the dendrite due to the influence of the matrix orientation hkl , and this angle is the angle to be finally determined, and the ratio of the temperature gradient at each grid point to the temperature gradient of the preferred crystal orientation can determine this angle.

[0099] The actual dendrite trunk growth rate V at the tip of the dendrite hkl and the temperature gradient G along the specific (hkl) /

[100] crystal orientation hkl are as follows:

[0100]

[0101]

[0102] Among them is the angle between the normal n of the molten pool boundary and the specific (hkl) /

[100] preferred crystal orientation.

[0103] The specific (hkl) /

[100] mentioned above means that during the rapid solidification of the molten pool, the columnar dendrites preferentially grow along these six <100> directions.

[0104] Arrange and solve the above formula for G n / V:

[0105]

[0106] Among them:

[0107] hkl refers to the crystal plane index of nickel-based single crystal;

[0108] G n is the temperature gradient;

[0109] V b is the laser scanning speed;

[0110] G x is the component of the temperature gradient on the X-axis.

[0111] On the right side of the equation, only the deviation angle of the preferred orientation dendrite is unknown. n , V b , G x Input, ψ hkl As an unknown parameter, it is uniformly valued and needs to be valued in a specific preferred direction, that is, <001> The six dendrite growth directions of the crystal plane family are output to obtain the G n / V collection.

[0112] The six directions of G n / V and K CET Compare and find the direction that can grow into a single crystal, compare its size, and intelligently select one of the 6 directions of G n / V, the judgment basis is to require the absolute value of G to be the largest, then at this time it is determined to be the optimal growth direction, corresponding to the optimal angle ψ hkl .

[0113] Then, the experimental data set of laser cladding of the nickel-based single crystal high-temperature alloy is automatically obtained through big data, and the laser cladding angle μ and the dendrite growth deviation angle ψ are constructed through the experimental data set. hkl The mapping relationship dataset is used as the first training dataset.

[0114] Then, a mapping relationship data set between the laser cladding angle μ and the single crystal area percentage is constructed through the experimental data set, and the mapping relationship data set is used as the second training data set.

[0115] 2) Determine the threshold and filter the training data set according to the threshold

[0116] K CET As a first reference value, G n / V and K CET By comparison, we can get G that can be grown into a single crystal. n / V collection.

[0117]

[0118] Determine the deviation threshold of the dendrite growth deviation angle according to the performance parameters, and each G in the set n / V corresponds to a dendrite growth deviation angle ψ hkl .

[0119] The performance parameters selected are to determine K CET Parameter value of a(m -1 s -1 ) and n are material related parameters, N0 is the nucleation density in the liquid (m -3 ), parameter φ c It is defined as the critical value of the area fraction in front of the premature solidification front consisting of newly nucleated grains.

[0120] The deviation threshold is the above G n / V The collection range corresponding to the collection.

[0121] The first training data set and the second training data set are screened based on the above deviation threshold to obtain a training data set for laser cladding, while data outside the deviation threshold is eliminated.

[0122] 3) Process the data through BP neural network to obtain the optimal laser cladding angle

[0123] The calculated dendrite growth deviation angle of the optimal growth direction and the training data set are input into the neural network model, which performs optimization and outputs the preset laser cladding angle μ and the angle ψ between the molten pool boundary normal phase and the specific preferred crystal orientation. hkl , thereby obtaining the optimal laser cladding angle corresponding to the optimal growth direction as the optimal angle that meets the alloy cladding requirements.

[0124] Secondly, the present invention further determines the laser cladding trajectory of the nickel-based single crystal high-temperature alloy according to the above method of determining the laser cladding angle.

[0125] 1) Determine data, search data and form a training data set

[0126] First, the experimental data set of laser cladding of the nickel-based single crystal high-temperature alloy is automatically obtained through big data.

[0127] Secondly, a mapping relationship dataset between the molten pool size and the cladding trajectory is constructed using the experimental dataset, and the mapping relationship dataset is used as the third training dataset.

[0128] 2) Determine the threshold and filter the training data set according to the threshold

[0129] The deviation threshold of the dendrite growth deviation angle is determined according to the performance parameters, and the selected performance parameters are the performance parameters for determining the size of the molten pool:

[0130]

[0131] in:

[0132] The material properties melting temperature T, thermal conductivity k, and thermal diffusivity α are constants;

[0133] P is the laser power absorbed by the substrate, W;

[0134] V b is the laser scanning speed, mm / s;

[0135] T0 is the substrate temperature, K;

[0136] The size of the molten pool is determined by x, y, and z, mm.

[0137] Input material properties melting temperature T, thermal conductivity k, thermal diffusion coefficient α, laser power P absorbed by the substrate, laser scanning speed V b , substrate temperature T0, the output x value range is x min ~x max , the value range of y is y min ~y max , then the deviation threshold is a set within the range of values of x and y.

[0138] The third training data set is screened based on the deviation threshold to obtain a training data set for the laser cladding trajectory, while data outside the deviation threshold is eliminated.

[0139] 3) Process the data through BP neural network to obtain the optimal laser cladding trajectory

[0140] The calculated melt pool size, the preset laser cladding angle μ, the angle ψ between the melt pool boundary normal and the specific preferred crystal orientation hkl The shape and size of the substrate to be repaired and the training data set are input into the BP neural network model, and the neural network model outputs the optimal cladding trajectory.

[0141] The present invention ultimately performs laser cladding on the base material according to a preset optimal laser cladding trajectory and an optimal laser cladding angle.

[0142] According to conventional processes, the base material needs to be pretreated before laser cladding, including grinding, polishing, cleaning, solid solution homogenization treatment and drying.

[0143] The present invention has no special limitation on the specific operation of the solution homogenization treatment, and the solution homogenization treatment process well-known to those skilled in the art can be adopted. In the present invention, the surface of the substrate can be polished with 600-grit silicon carbide paper and cleaned in methanol, and a DD6 standard solution is selected for homogenization at 1300 °C for 3 h.

[0144] Then, according to the preset laser cladding angle, the included angle between the laser incident direction and the nickel-based single crystal superalloy substrate is adjusted, and according to the preset laser cladding track, the nickel-based single crystal superalloy after solution treatment is laser-clad by means of coaxial powder feeding, and the nickel-based single crystal superalloy is repaired layer by layer.

[0145] The present invention can also perform heat treatment on the nickel-based single crystal superalloy after laser cladding. It includes a dissolution step in the first step and one or two aging steps thereafter.

[0146] The dissolution step therein is to heat the sample to 1120 °C and cool for 2 h, and then perform water quenching. The first aging step is to heat the sample to 1000 °C and heat for 2 h, and then perform furnace cooling. The second aging step is to heat the sample to 788 °C and cool for 8 h, and then perform furnace cooling. Examples

[0147] Example 1

[0148] Take a nickel-based single crystal turbine blade with a substrate material of DD5. Before repairing the blade, the blade needs to be inspected. The non-destructive testing method is used to find the repair part, determine the substrate orientation and cut the repair part.

[0149] The surface of the blade to be repaired after polishing is cleaned with acetone alcohol.

[0150] Only the repair steps for the first layer are described here. The following repair layers are gradually repaired according to the steps of the first repair layer, and a layer of crystal structure with fine grains and consistent orientation is obtained.

[0151] Use K438 powder to repair DD5 nickel-based single crystal. The process parameters for the first layer are set as follows: laser power 400 W, spot diameter 2 mm, scanning speed 1 mm / s, powder feeding rate 10 g / min, and coaxial powder feeding is used for cladding.

[0152] The material properties are melting temperature T = 1675 K, thermal conductivity k = 0.0332 kj / (m·s·k), thermal diffusivity α = 3.6×10 -9 , P = 400 W, V b = 1 mm / s, substrate temperature T0 = 300 K, n = 3.4. Input into the BP neural network to obtain the experimental data set and the threshold range. After continuous training by the BP neural network, output ψhkl = 26.7°, μ = 15.9°.

[0153] The melting temperature of the material properties is T = 1675K, the thermal conductivity is k = 0.0332 kj / (m·s·k), and the thermal diffusivity is α = 3.6×10 -9 , P = 400W, V b = 1mm / s, the substrate temperature is T0 = 300k, n = 3.4, ψ hkl = 26.7°, μ = 15.9°. The blade geometry profile and size are input into the BP neural network to obtain the experimental data set and the threshold range. After continuous training by the BP neural network, the preset cladding trajectory is output: two circles of cladding are performed on the outer layer according to the profile, and laser cladding repair is performed on the inner layer at an angle of 45° with the x-axis direction.

[0154] Laser cladding is performed according to the preset cladding trajectory and laser cladding angle output above. During the cladding process, the cladding trajectory is monitored and reported in real time to play the role of manual detection.

[0155] Comparative Example 1

[0156] The method of laser directed cladding of nickel-based single crystal based on artificial intelligence in Example 1 is not adopted. The same substrate material and cladding material as in Example 1 are used. According to the same process parameters, the laser power is set to 400W, the spot diameter is 2mm, the scanning speed is 1mm / s, the powder feeding rate is 10g / min, and coaxial powder feeding is used for cladding. There is no laser cladding angle, and the laser cladding trajectory is to perform two circles of cladding on the outer layer according to the profile, and laser cladding repair is performed on the inner layer along the x-axis direction.

[0157] Figure 1 and Figure 2 respectively show the scanning electron microscope images of the cross-section of the cladding coating of the substrate material after laser cladding in Example 1 and Comparative Example 1.

[0158] From Figure 1 and Figure 2 By comparison, it can be seen that more dendritic single crystal areas grow epitaxially using the method of the present invention.

[0159] The above embodiments of the present invention do not elaborate on all details and do not limit the present invention to only the above-described embodiments. Various changes, modifications, substitutions, and variations made to these embodiments by those of ordinary skill in the art without departing from the principles and purposes of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for laser directional cladding of nickel-based single crystals based on artificial intelligence, which controls the growth of the nickel-based single crystal along the correct epitaxial growth direction, i.e., the [001] dendrite growth direction of the nickel-based single crystal, according to artificial intelligence, comprising: 1) Determine the laser cladding angle of nickel-based single crystal high-temperature alloy based on artificial intelligence: 1) Determine data, search data and form a training data set according to the following formula: in: The material properties melting temperature T, thermal conductivity k, and thermal diffusivity α are constants; P is the laser power absorbed by the substrate, W; V b is the laser scanning speed, mm / s; T0 is the substrate temperature, K; The size of the molten pool is determined by x, y, and z, mm; Input material properties: melting temperature T, thermal conductivity k, thermal diffusivity α, laser power P absorbed by the substrate, laser scanning speed V b , substrate temperature T0, then the molten pool size and temperature gradient G can be output according to the above formula n and its components G x , G y , G z ; Then determine G according to the following formula n / V, where the value of the characteristic n of the nickel-based single crystal material is 3.4: in: hkl refers to the crystal plane index of nickel-based single crystal; G n is the temperature gradient; V b is the laser scanning speed; G x is the component of the temperature gradient on the X-axis; On the right side of the equation, only the deviation angle of the dendrites with preferred orientation is unknown. Substitute the above G n , V b , G x as inputs, take ψ hkl as an unknown parameter and uniformly assign values, and it is necessary to assign values in a specific preferred direction, that is, the 6 dendrite growth directions of the <001> crystal plane family, and output the G n / V sets in 6 directions; Intelligently screen out one of the six directions of G n / V. The judgment basis is that the absolute value of G is required to be the largest. Then the determined direction at this time is the optimal growth direction; Automatically obtain an experimental data set of laser cladding of the nickel-based single crystal high-temperature alloy through big data; Construct a mapping relationship dataset between the laser cladding angle μ and the dendrite growth deviation angle ψ through the experimental dataset hkl and use the mapping relationship dataset as the first training dataset; A mapping relationship dataset between the laser cladding angle μ and the single crystal area percentage is constructed through the experimental dataset, and the mapping relationship dataset is used as the second training dataset; 2) Determine the threshold and filter the training data set according to the threshold; 3) Process the data through BP neural network to obtain the optimal laser cladding angle; 2) Determine the laser cladding trajectory of nickel-based single crystal high-temperature alloy based on artificial intelligence: 1) Determine data, search data and form a training data set; 2) Determine the threshold and filter the training data set according to the threshold; 3) Process the data through BP neural network to obtain the optimal laser cladding trajectory; 3) Presetting an optimal laser cladding trajectory and laser cladding the pretreated substrate material at an optimal laser cladding angle: Presetting a laser cladding angle, adjusting the angle between the laser incident direction and the nickel-based single crystal high-temperature alloy substrate, and laser cladding the nickel-based single crystal high-temperature alloy according to the preset laser cladding trajectory; The above method is based on setting the positive direction of the y-axis as the laser scanning direction and the z-axis direction as the [001] direction.

2. The method for laser direct cladding of nickel-based single crystals based on artificial intelligence according to claim 1, characterized in that In the artificial intelligence-based determination of the laser cladding angle of the nickel-based single crystal high-temperature alloy, a threshold is determined and a training data set is screened according to the threshold in the following manner: Taking K CET as the first reference value, compare G n / V with K CET to obtain the set of G n / V that can grow into single crystals. Determine the deviation threshold of the dendritic growth deviation angle according to the performance parameters, and each G n / V in the set corresponds to a dendritic growth deviation angle ψ hkl , The selected performance parameters are for determining the parameter values of K CET where a and n are material-related parameters, the unit of a is m -1 s -1 , N0 is the nucleation density in the liquid, the unit is m -3 , and the parameter φ c is defined as the critical value of the area fraction in front of the prematurely solidified front composed of newly nucleated grains The deviation threshold is the set range corresponding to the above G n / V set, The first training data set and the second training data set are screened based on the above deviation threshold to obtain a training data set for laser cladding, while data outside the deviation threshold is eliminated.

3. The method for laser direct cladding of nickel-based single crystal based on artificial intelligence according to claim 1, characterized in that In the determination of the laser cladding angle of nickel-based single crystal superalloys based on artificial intelligence, the optimal laser cladding angle obtained by processing data through a BP neural network is to input the calculated dendrite growth deviation angle of the optimal growth direction and the training data set into the neural network model, which is optimized by the neural network model to output the preset laser cladding angle μ, as well as the angle ψ between the normal phase of the molten pool boundary and a specific preferred crystal orientation. hkl .

4. The method for laser directed energy deposition of nickel-based single crystals based on artificial intelligence according to claim 1, characterized in that The method of determining the determined data in the laser cladding trajectory of the nickel-based single crystal high-temperature alloy based on artificial intelligence, searching the data and forming the training data set is to automatically obtain the experimental data set for laser cladding of the nickel-based single crystal high-temperature alloy through big data; constructing a mapping relationship data set between the molten pool size and the cladding trajectory through the experimental data set, and using the mapping relationship data set as the third training data set.

5. The method for laser direct cladding of nickel-based single crystal based on artificial intelligence according to claim 1, characterized in that The artificial intelligence-based determination of the threshold value in the laser cladding trajectory of the nickel-based single crystal high-temperature alloy and the screening of the training data set according to the threshold value is to determine the deviation threshold value of the dendrite growth deviation angle according to the performance parameter, and the selected performance parameter is the performance parameter for determining the molten pool size: in: The material properties melting temperature T, thermal conductivity k, and thermal diffusivity α are constants; P is the laser power absorbed by the substrate, W; V b is the laser scanning speed, in mm / s; T0 is the substrate temperature, K; The size of the molten pool is determined by x, y, and z, mm; Input material properties: melting temperature T, thermal conductivity k, thermal diffusivity α, laser power P absorbed by the substrate, laser scanning speed V b , substrate temperature T0, the value range of the output x is x min ~x max , the value range of y is y min ~y max , then the deviation threshold is the set within the value ranges of x and y; The third training data set is screened based on the deviation threshold to obtain a training data set for the laser cladding trajectory, while data outside the deviation threshold is eliminated.

6. The method for laser directed energy deposition of nickel-based single crystals based on artificial intelligence according to claim 1, wherein The optimal laser cladding track obtained by processing data through a BP neural network in the laser cladding track of nickel-based single crystal superalloys determined based on artificial intelligence is the calculated molten pool size, the preset laser cladding angle μ, the angle ψ between the normal direction of the molten pool boundary and a specific preferred crystal orientation hkl , the shape profile size of the substrate to be repaired, and the training data set are input into the BP neural network model, and the optimal cladding track is output by the neural network model.

7. The method for laser direct cladding of nickel-based single crystal based on artificial intelligence according to claim 1, characterized in that The base material is pretreated before laser cladding, including grinding, polishing and cleaning the base surface.

Images