阅读说明：本技术 积层制造方法 (Method for manufacturing build-up layer ) 是由 陈红章 罗裕龙 杨浩青 郑芳田 于 2021-06-24 设计创作，主要内容包括：本发明提供一种积层制造方法。此方法包含在粉层上进行激光粉床熔融(L-PBF)制程。然后,获得L-PBF制程后的粉层的第一表面粗糙度,以产生第一表面轮廓。使用吸收度及一组重熔制程参数数据来进行热传模拟。通过第一表面轮廓及低通滤波器,来获得激光重熔后的粉层的第二表面轮廓。然后,重复调整此组重熔制程参数的数值,以进行热传模拟,直至自第二表面轮廓所预测的第二表面粗糙度值小于或等于表面粗糙度阈值,借以获得重熔制程参数的最佳数值,以进行重熔制程来减少L-PBF制程后的粉层的表面粗糙度。(The invention provides a method for manufacturing an additive layer. The method includes performing a laser powder bed fusion (L-PBF) process on the powder layer. Then, a first surface roughness of the powder layer after the L-PBF process is obtained to generate a first surface profile. The heat transfer simulation is performed using the absorbance and a set of reflow process parameter data. And obtaining a second surface profile of the laser-remelted powder layer through the first surface profile and the low-pass filter. And then, repeatedly adjusting the numerical values of the set of remelting process parameters to perform heat transfer simulation until the second surface roughness value predicted from the second surface profile is less than or equal to the surface roughness threshold value, so as to obtain the optimal numerical value of the remelting process parameters, and perform the remelting process to reduce the surface roughness of the powder layer after the L-PBF process.)

1. A method of manufacturing an additive layer, comprising:

providing a powder bed, wherein the powder bed comprises a base material and a powder layer, and the powder layer is arranged on the base material;

performing a laser powder bed fusion (L-PBF) process on the powder layer using a set of melting parameter data, wherein the set of melting parameter data includes a first laser power value, a first scanning speed value and a first scanning space (hooking space) value;

obtaining a set of property data relating to the powder bed, wherein the set of property data includes a plurality of material properties and a plurality of optical properties of a top surface of the powder layer;

obtaining a first surface roughness value of the top surface of the powder layer after the laser powder bed melting process;

generating a first surface profile from the first surface roughness value based on a Gaussian probability assumption (Gaussian probability assumption);

obtaining an absorption of the powder layer according to a Bidirectional Reflection Distribution Function (BRDF) using the first surface profile and the set of property data;

providing a set of remelting process parameter data to perform a remelting process on the powder layer, wherein the set of remelting process parameter data includes a second laser power value, a second scanning speed value and a second scanning interval value;

performing a heat transfer simulation operation by using the set of remelting process parameter data, the set of property data, and the absorbance of the powder layer to simulate the remelting process, thereby obtaining a temperature distribution and an average melting time in a molten pool area;

calculating a cut-off frequency (cut-off frequency) by using the average melting time and the set of property data, thereby obtaining a low pass filter (low pass filter);

obtaining a second surface profile of the top surface of the powder layer after the laser remelting process by using the first surface profile and the low-pass filter;

predicting a second surface roughness value of the top surface of the powder layer after the laser remelting process based on the second surface profile;

repeatedly adjusting the set of remelting process parameter data to perform the heat transfer simulation operation until the second surface roughness value is less than or equal to a surface roughness threshold value; and

after the laser powder bed melting process, a laser remelting process is performed on the top surface of the powder layer with the set of remelting process parameter data.

2. The additive manufacturing method of claim 1, wherein the first surface roughness value is obtained by a virtual metrology method.

3. The additive manufacturing method of claim 2, wherein the virtual measurement method is based on a Neural Network-like algorithm, a Back Propagation Neural Network (BPNN) algorithm, a General Regression Neural Network (GRNN) algorithm, a Radial Basis Function Neural Network (RBFNN) algorithm, a Simple Regression Network (SRN), a Support Vector Data Description (SVDD) algorithm, a Support Vector Machine (Mac Vector) algorithm, a Multiple Regression (MR) algorithm, a Partial Least Squares (PLS) algorithm, a Nonlinear substitution biased minimum Squares (GLLS) algorithm, or a Generalized model Least Squares (GLLS) algorithm.

4. The method of claim 1, wherein the step of predicting the second surface roughness value comprises:

converting the first surface profile from a spatial domain to a frequency domain by a Fast Fourier Transform (FFT) algorithm;

applying the low-pass filter to the first surface profile of the frequency domain to obtain the second surface profile of the frequency domain;

transforming the second surface profile from the frequency domain to the spatial domain by a fast fourier transform algorithm; and

calculating the second surface roughness value from the second surface profile of the spatial domain.

5. The additive manufacturing method of claim 1, further comprising:

obtaining a plurality of melting times corresponding to a plurality of nodes in the molten bath region by using the temperature profile; and

the average melting time is calculated by using the plurality of melting times and the plurality of nodes.

6. The additive manufacturing method of claim 1 wherein the thermal transfer simulation operation is a finite element thermal transfer simulation operation.

7. A method of performing additive manufacturing, the method comprising:

obtaining a plurality of first surface roughness values of at least one powder layer, wherein the at least one powder layer has been processed by a laser powder bed fusion process using a set of fusion parameter data, and the set of fusion parameter data includes a first laser power value, a first scan velocity value, and a first scan pitch value;

performing a simulation operation on each of the first surface roughness values of each of the at least one powder layer to obtain a reflow operation table, wherein the reflow operation table includes a plurality of sets of reflow process parameter data corresponding to the plurality of first surface roughness values in a one-to-one manner, and the simulation operation includes:

obtaining a set of property data relating to one of the at least one powder layer, wherein the set of property data comprises a plurality of material properties and a plurality of optical properties of each of the at least one powder layer;

generating a first surface profile from one of the first surface roughness values based on a gaussian probability hypothesis;

obtaining an absorbance of one of the at least one powder layer according to a bi-directional reflectance distribution function using the first surface profile and the set of property data;

providing a set of remelting process parameter data to perform a remelting process on the powder layer, wherein the set of remelting process parameter data includes a second laser power value, a second scanning speed value and a second scanning interval value;

performing a heat transfer simulation operation by using the set of reflow process parameter data, the set of property data and the absorbance of one of the at least one powder layer to simulate the reflow process, thereby obtaining a temperature distribution and an average melting time in a molten pool area;

calculating a cutoff frequency by using the average melting time and the set of property data to obtain a low pass filter;

obtaining a second surface profile of one of the at least one powder layer by using the first surface profile and the low pass filter; and

predicting a second surface roughness value of one of the at least one powder layer based on the second surface profile; and

repeatedly adjusting the set of remelting process parameter data to perform the heat transfer simulation operation until the second surface roughness value is less than or equal to a surface roughness threshold value;

obtaining another roughness value of a top surface of another powder layer after performing the laser powder bed fusion process on the top surface of the another powder layer; and

and finding another set of remelting process parameter data in the remelting operation table according to the another roughness value, so as to perform a laser remelting process on the top surface of the another powder layer by using another set of remelting process parameter data after the laser powder bed melting process.

8. The additive manufacturing method of claim 7 wherein the first surface roughness values are on the same powder layer.

9. The additive manufacturing method of claim 7 wherein the first surface roughness values are on different powder layers.

10. The additive manufacturing method of claim 7, wherein the another roughness value is obtained by a virtual metrology method.

Technical Field

The present invention relates to an Additive Manufacturing (AM) method, and more particularly, to an AM method for performing laser-remelting (laser-remelting) process with an optimum value of a remelting process parameter.

Background

Additive Manufacturing (AM), commonly known as 3D printing, is a technique of creating a digital computer model file, heating metal powder or plastic material to make it in a molten and plastic state, and then stacking the metal powder or plastic material layer by layer to obtain a workpiece. In recent years, build-up process technology has grown rapidly because of its ability to fabricate functional components with highly complex structures in a fast, versatile and cost-effective manner that minimizes waste of metal powder. Laser powder bed fusion (L-PBF) process is one of the most common additive manufacturing techniques.

In the laser powder bed fusion process, 3D devices with complex geometries are fabricated layer by using a control laser beam to selectively fuse specific areas of the metal powder bed. However, due to the random nature of the laser powder bed melting process, the top surface roughness of each solidified layer (solidified layer) may vary even when optimized process conditions are used at different locations on the fabrication plate. As such, the mechanical properties of the fabricated device frequently change as the device changes. Thus, laser polishing (laser polishing) techniques or laser re-melting processes are used to reduce the surface roughness of each cured layer.

However, the parameters of the laser reflow process for each layer are usually determined by a trial and error method (trial-and-error). In view of the above, it is desirable to provide a method for manufacturing a build-up layer, which performs a laser reflow process with optimized reflow process parameters.

Disclosure of Invention

An object of the present disclosure is to provide a method for manufacturing a build-up layer, which can obtain optimized reflow process parameters by simulation, so as to effectively perform a laser reflow process to reduce the surface roughness of a powder layer after a laser powder bed melting process.

In accordance with the above objectives, an aspect of the present disclosure provides a method for manufacturing an integrated circuit. In the lamination manufacturing method, first, a powder bed is provided, wherein the powder bed comprises a base material and a powder layer arranged on the base material. Then, a laser powder bed fusion (L-PBF) process is performed on the powder layer using a set of fusion parameter data. The set of melting parameter data includes a first laser power value, a first scanning speed value and a first scanning space (hooking space) value. A set of property data about the powder bed is obtained. The set of property data includes material properties and optical properties of the top surface of the powder layer. Then, a first surface roughness value of the top surface of the powder layer after the laser powder bed melting process is obtained, and the first surface profile is generated from the first surface roughness value based on Gaussian probability assumption (Gaussian probability assumption). Subsequently, the absorption of the powder layer is obtained according to a Bidirectional Reflection Distribution Function (BRDF) by using the first surface profile and the set of property data. A set of reflow process parameter data is provided to perform a reflow process on the powder layer. The set of reflow process parameter data includes a second laser power value, a second scan speed value and a second scan spacing value. Then, a heat transfer simulation operation is performed by using the set of reflow process parameter data, the set of property data and the absorptivity of the powder layer to simulate the reflow process, so as to obtain the temperature distribution and the average melting time in the molten pool area. By using the average melting time and the set of property data, a cut-off frequency (cut-off frequency) is calculated, thereby obtaining a low pass filter (low pass filter). Then, a second surface profile of the top surface of the powder layer after the laser remelting process is obtained by using the first surface profile and the low-pass filter. Then, a second surface roughness value of the top surface of the powder layer after the laser remelting process is predicted based on the second surface profile. The reflow process parameter data is adjusted repeatedly to perform the heat transfer simulation operation until the second surface roughness value is less than or equal to the surface roughness threshold value. And then, performing a laser remelting process on the top surface of the powder layer after the laser powder bed melting process according to the set of remelting process parameter data.

According to an embodiment of the present disclosure, the first surface roughness value is obtained by a virtual measurement method.

According to an embodiment of the disclosure, the virtual measurement method is based on a Neural Network-like algorithm, a Back Propagation Neural Network (BPNN) algorithm, a General Regression Neural Network (GRNN) algorithm, a Radial Basis Function Neural Network (rbn) algorithm, a Simple Regression Network (SRN), a Support Vector Data Description (SVDD) algorithm, a Support Vector Machine (SVM) algorithm, a Multiple Regression (MR) algorithm, a Partial Least square (Partial Least square) algorithm, a Nonlinear substitution biased minimum Squares (PLS) algorithm, or a Nonlinear substitution biased minimum Squares (glls) algorithm.

According to an embodiment of the present disclosure, the step of predicting the second surface roughness value includes converting the first surface profile from a spatial domain to a frequency domain by a Fast Fourier Transform (FFT) algorithm. Then, a low-pass filter is applied to the first surface profile of the frequency domain to obtain a second surface profile of the frequency domain. The second surface profile is then transformed from the frequency domain to the spatial domain by a fast Fourier transform algorithm. Then, a second surface roughness value is calculated from the second surface profile of the spatial domain.

According to an embodiment of the present disclosure, the method further includes obtaining a plurality of melting times corresponding to a plurality of nodes in the molten bath area by using the temperature profile. Then, by using the melting time and such nodes, an average melting time is calculated.

According to an embodiment of the present disclosure, the thermal transfer simulation operation is a finite element thermal transfer simulation operation.

According to another aspect of the present disclosure, a method for performing a laser reflow process is provided. The method includes obtaining a plurality of first surface roughness values of at least one powder layer, wherein the at least one powder layer has been processed by a laser powder bed fusion process using a set of fusion parameter data. The set of melting parameter data includes a first laser power value, a first scan speed value and a first scan pitch value. Then, a simulation operation is performed on each first surface roughness value of each at least one powder layer, so as to obtain a reflow operation table. The reflow operation table comprises a plurality of sets of reflow process parameter data corresponding to the first surface roughness values in a one-to-one manner. The simulation operation includes obtaining a set of property data regarding one of the at least one powder layers, wherein the set of property data includes material properties and optical properties of each of the at least one powder layers. The first surface profile is generated from one of the first surface roughness values based on a gaussian probability hypothesis. Then, the absorbance of one of the at least one powder layer is obtained according to the bi-directional reflection distribution function by using the first surface profile and the set of property data. Providing a set of re-melting process parameter data to perform a re-melting process on the powder layer, wherein the set of re-melting process parameter data includes a second laser power value, a second scanning speed value and a second scanning interval value. Then, a heat transfer simulation operation is performed by using the set of remelting process parameter data, the set of property data and the absorbance of at least one powder layer to simulate the remelting process, so as to obtain the temperature distribution and the average melting time in the molten pool area. The low pass filter is obtained by calculating the cutoff frequency using the average melting time and the set of property data. Then, a second surface profile of one of the at least one powder layer is obtained by using the first surface profile and the low pass filter. A second surface roughness value of one of the at least one powder layer is predicted based on the second surface profile. The simulating operation includes repeatedly adjusting the set of reflow process parameter data to perform the heat transfer simulating operation until the second surface roughness value is less than or equal to the surface roughness threshold value. After the simulation operation, another roughness value of the top surface of the other powder layer is obtained after performing a laser powder bed fusion process on the top surface of the other powder layer. And finding another set of remelting process parameter data in the remelting operation table through another roughness value, so as to perform the laser remelting process on the top surface of another powder layer after the laser powder bed melting process by using another set of remelting process parameter data.

According to an embodiment of the present disclosure, the first surface roughness values are on the same powder layer.

According to an embodiment of the present disclosure, the first surface roughness values are on different powder layers.

According to an embodiment of the present disclosure, the roughness value is obtained by a virtual measurement method.

Therefore, with the disclosed additive layer manufacturing method, the quality of the workpiece processed by laser powder bed fusion can be optimized by controlling the surface roughness of each layer.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.

Drawings

The aspects of the present disclosure are best understood from the following detailed description when read with the accompanying drawings. It is noted that, as is standard practice in the industry, many features are not drawn to scale. In fact, the dimensions of many of the features may be arbitrarily scaled for clarity of discussion.

FIGS. 1A and 1B are flow diagrams illustrating methods of additive layer manufacturing according to some embodiments of the present disclosure;

FIG. 2 is a schematic diagram illustrating a build-up process tool on a process tool according to some embodiments of the present disclosure;

FIGS. 3A and 3B illustrate simulated surface profiles of different average surface roughness in accordance with some embodiments of the present disclosure;

fig. 4A and 4B illustrate the interaction between a powder layer having a perfectly smooth surface and a rough surface and laser energy, respectively;

FIG. 5 depicts surface profiles before and after laser reflow in accordance with some embodiments of the present disclosure;

FIG. 6 is a graph showing a surface roughness value before laser reflow, an experimental surface roughness value after a laser reflow process with different sets of laser reflow parameter data, and a simulated surface roughness value, in accordance with some embodiments of the present disclosure;

FIG. 7 is a flow chart illustrating a method of additive manufacturing according to some embodiments of the present disclosure;

FIG. 8 is a block diagram illustrating an intelligent additive manufacturing architecture according to some embodiments of the present disclosure;

FIG. 9 is a flow chart illustrating the workflow of an automated virtual metrology system according to some embodiments of the present disclosure.

[ notation ] to show

100 method

110,115,120,125,130,135,140,150,160,170,180,185,190 operation

200 device

202 laser

204 laser beam

210 powder bed

212 powder layer

214 base material

216 powder bed container

500 method

510,520,530,540,550: operation

600 intelligent build-up manufacturing framework

610 build-up manufacturing machine

612 controller

614 build-up manufacturing equipment

620 in-situ measurement system

630 external measurement system

640 automated virtual metrology system

650 simulator for additive manufacturing

660 intelligent compensator

670 tracking planner

710,720,730,740,750,760 operation

CL coaxial light

DB database

EM external measurement data

IM-in-situ metrology data

IS in situ sensing data

PA_fAdjustment value of external process parameters

PA_nOn-board process parameter adjustment

PP Process parameters

PR recommended range of process parameters

PT tracking value of process parameter

VM virtual metrology data

Detailed Description

The embodiments of the present invention will be described in detail with reference to the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings and the description to refer to the same or like parts.

Although current systems are available to monitor the occurrence of defects such as void formation and excessive surface roughness in individual fused powder layers during a laser powder bed fusion process, there is a need for a method that can be combined with such systems to positively modify the powder layers of a laser powder bed fusion process.

Referring to fig. 1A, 1B and 2, fig. 1A and 1B are flow diagrams illustrating a method 100 of additive layer manufacturing according to some embodiments of the present disclosure, and fig. 2 is a schematic diagram illustrating an additive layer processing tool 200 on a processing tool according to some embodiments of the present disclosure. In operation 110, a powder bed and a laser are provided. In some embodiments, as shown in fig. 2, the apparatus 200 comprises a powder bed 210 within a powder bed container 216, wherein the powder bed 210 comprises a plurality of powder layers 212 deposited on a substrate 214. In addition, each powder layer 212 includes a plurality of powder materials.

Next, in operation 115, a laser powder bed fusion (L-PBF) process is performed on the top surface of the powder layer 212 by using a laser 202 having a set of fusion parameter data, as shown in fig. 2. In some embodiments, the set of melting parameter data includes a first laser power value, a first scan speed value, and a first pitch (hooking space) value. In some embodiments, after the powder layer 212 is irradiated by the laser 202, the powder layer 212 is melted and solidified, and thus a solidified layer may be obtained. In some embodiments, there are multiple powder layer regions on the substrate. During the laser powder bed fusion process on the powder layer 212, the laser 202 is used to provide the laser beam 204 to the respective powder layer region of the powder bed 210, and the powder layer 212 is fused and formed by controlling specific parameters to obtain the desired profile of the product. As the powder in the respective powder layer regions is melted, a plurality of melt pools are formed on the powder bed 210.

Then, an operation 120 is performed to obtain a set of property data regarding the powder bed (e.g., powder bed 210). In some embodiments, the set of property data includes material properties and optical properties of a top surface of the powder layer. In some embodiments, the material properties include, but are not limited to, thermal conductivity, density, specific heat, solidus temperature, and liquidus temperature. In some embodiments, the optical properties include, but are not limited to, the refractive index of the powder material for different laser wavelengths.

Referring to fig. 1A, operation 125 is performed to obtain a first surface roughness value of the top surface of the powder layer after the laser powder bed melting process. In some embodiments, the first surface roughness value is obtained by using a virtual metrology method. In some embodiments, the virtual measurement method is based on a Neural Network-like algorithm, a Back Propagation Neural Network (BPNN) algorithm, a General Regression Neural Network (GRNN) algorithm, a Radial Basis Function Neural Network (RBFNN) algorithm, a Simple Regression Network (SRN), a Support Vector Data Description (SVDD) algorithm, a Support Vector Machine (SVM) algorithm, a Multiple Regression (MR) algorithm, a Partial Least Squares (PLS) algorithm, a Nonlinear substitution biased minimum Squares (GLLS) algorithm, or a Generalized model Neural Network (GLNN) algorithm. In some embodiments, the first surface roughness value is obtained by using a surface roughness tester.

An operation 130 is performed to construct a first surface profile from the first surface roughness values obtained in operation 125. In some embodiments, operation 130 is based on a Gaussian probability assumption (Gaussian probability assumption). For example, fig. 3A and 3B show simulated profiles with average surface roughness values Ra 2.49 μm and Ra 5.07 μm, respectively, in accordance with some embodiments of the present disclosure. In addition, one method to generate the surface profile is to model a random rough surface with a one-dimensional function, which is described in a statistical fashion by two distribution functions (height probability distribution and auto-dependent function). The height probability distribution describes the height of the surface from a particular average reference value, while the self-dependent function describes the variance of the aforementioned height laterally along the surface. In some embodiments, equation (1) is a common model, with an approximate height probability distribution as a gaussian function.

In formula (1), σ is a root mean square (rms) height (surface roughness Ra), which is equal to the standard deviation, and ζ is a height function. In some embodiments, the self-dependent function may be defined as equation (2):

in formula (2), x₁And x₂At two different points along the surface and τ is the correlation length.

The present disclosure is incorporated by reference into "D".J.Powell,and A.Kaplan,"A ray-tracing analysis of the absorption of light by smooth and rough metal surfaces, "Journal of applied physics, vol.101, p.113504, 2007".

After the first surface profile is generated, it is assumed that the absorption may be different when the laser is irradiated onto surfaces having different surface roughness values. Then, an operation 135 is performed to obtain the absorbance of the powder bed according to a Bidirectional Reflection Distribution Function (BRDF). In some embodiments, the first surface profile and the set of property data are used to obtain the absorbance. When a laser powder bed melting process or a laser remelting process is performed on the powder bed, the laser beam is irradiated on the surface of the solidified layer, part of the energy of the laser beam is absorbed by the metal powder (powder material), and the rest of the energy is reflected. Typically, the absorbed laser energy is significantly dependent on the surface roughness of the powder bed and the optical properties of the metal powder (powder material). In particular, for rough surfaces (e.g., having a large surface roughness value), the degree of interaction between the laser beam and the surface morphology increases dramatically, so the laser energy absorbed by the powder bed increases. Therefore, the influence of the surface morphology on the laser absorption must be taken into account.

The two-way reflection distribution function algorithm for operation 135 may be equation (3).

In the formula (3), θ_sIs the angle of reflection;andincident light energy and reflected light energy respectively; and omega_iAnd omega_sIncident angle and solid angle, respectively. Integrating equation (3) over the entire sphere yields a hemispherical reflectivity with the direction:

then, the absorbance function of the surface is obtained as formula (5):

A′_λ(Ω_i)＝1-ρ′_λ(Ω_i) (5)

it can be appreciated that the absorption increases significantly with increasing surface roughness, since higher roughness results in a larger number of "hilly" and "valley" structures, which result in more interaction between the laser energy and the surface, as shown in fig. 4A and 4B, which illustrate the interaction between the laser energy and a powder layer having a perfectly smooth surface and a rough surface, respectively. For example, when the surface roughness is 2.5 μm, the calculated absorbance is about 0.35; when the surface roughness was 4.98 μm, the calculated absorbance was about 0.41.

Referring to fig. 1B, an operation 140 is performed to provide a set of reflow process parameter data of the laser to perform a reflow process on the powder layer. In some embodiments, the set of reflow process parameter data includes a second laser power value, a second scan speed value and a second scan pitch value.

In operation 150, a heat transfer simulation operation is performed. In some embodiments, the heat transfer simulation operation is a finite element heat transfer simulation operation to simulate a remelting process to obtain a temperature distribution in the molten bath. In some embodiments, the set of property data (e.g., obtained in operation 120), the absorption of the powder layer (e.g., obtained in operation 135), and the set of reflow process parameter data (e.g., obtained in operation 140) are employed to perform a thermal transfer simulation operation.

In some embodiments, the control equation for heat transfer in the molten bath of the powder bed may be written as:

where ρ represents the density of the material (kg/m)³) C represents specific heat (J/kg-K), T represents temperature (K), and Q represents the amount of heat generated per unit volume (W/m)³). At one endIn one embodiment, the powder layer and the substrate are assumed to have an initial temperature of 293K. Meanwhile, the boundary conditions of heat transfer on the top surface are set as:

in formula (7), k_nDenotes the thermal conductivity in the direction perpendicular to the surface, h_cRepresenting heat transfer coefficient (e.g., h)_c＝100W/m²-K)，T_aIndicating normal temperature (e.g. T)_a293K), σ denotes Stefan constant (Stefan constant) (which is 5.669 × 10^-8W/m²K) and epsilon represents the emissivity (e.g. 0.4 for solid and 0.1 for molten material). In particular, h (T-T)_a) Is due to heat loss by convection, and σ ε (T)⁴-T_a ⁴) Due to heat loss from the radiation. Further, q is_laserRepresents the absorbed laser energy, and is represented by formula (8).

In the formula (8), P is laser energy, r₀Is the radius of the laser beam, v is the laser scanning speed, x, y and z are the coordinates of the simulation domain, and a is the calculated absorbance described above. It should be noted that the simulation takes into account both the phase change phenomenon, which can occur when the material changes from bulk to liquid and from liquid to gas, and the effect of Marangoni convection (Marangoni convection) on the formation of the melt pool due to anisotropic thermal conductivity.

Further, operation 150 is also performed to obtain an average melting time. It is understood that the melting time refers to the time during which the material remains in the liquid state. In some embodiments, the step of obtaining an average melting time comprises obtaining a plurality of melting times corresponding to nodes in a plurality of melt pools from the temperature profile. Then, the average melting time is calculated by using the melting time and the node, as shown in equation (9):

in formula (9), T_m ⁱIs the melting time of the individual nodes, and N is the total number of nodes in the melt pool.

In operation 160, a cut-off frequency (cut-off frequency) is calculated. In some embodiments, the set of property data (e.g., obtained by operation 120) and the average melting time (e.g., obtained by operation 150) are calculated according to equation (10) to calculate the cutoff frequency:

in formula (10), f_crDenotes the cut-off frequency (1/mm), and μ denotes the dynamic viscosity (Ns/m)²) ρ represents density (kg/m)³) And T is_mIndicating the melting time. It is understood that both dynamic viscosity and density are included in the set of property data. For example, table 1 shows the cut-off frequencies for three sets of laser reflow process parameter data.

TABLE 1

Laser energy (W)	Scanning speed (mm/s)	Cut-off frequency (Hz)
			150	600	6638.8
180	680	6394.3
			200	750	6488.8

The present disclosure refers to and incorporates the mathematical models set forth in "e.ukar, a.lamikiz, s.mart i niez, i.taberno, and l.l.de Lacalle," rough prediction on laser polarized surfaces, "Journal of Materials Processing Technology, vol.212, pp.1305-1313,2012".

Then, operation 170 is performed to obtain a second surface profile of the top surface of the powder layer after the laser remelting process. In some embodiments, the first surface profile and the low pass filter with a cut-off frequency are used to predict the second surface profile after the laser melting process. Referring to fig. 5, fig. 5 shows surface profiles before and after laser remelting, in which the surface profile before laser remelting is obtained after a laser powder bed melting process at a laser energy of 150W and a scanning speed of 600mm/s, and the surface profile after laser remelting is a simulated profile obtained after applying a low pass filter to a frequency domain. It was observed that the laser remelting process reduced the average surface roughness value from 4.5 μm to 2.8 μm. It is to be understood that the low pass filter may pass lower frequencies but block higher frequencies. Thus, the second surface profile is composed of a lower frequency signal.

Then, operation 180 is performed to predict a second surface roughness value of the top surface of the powder layer after the laser remelting process. The second surface roughness value is based on the second surface profile obtained in operation 170. In some embodiments, the step of predicting the second surface roughness value comprises converting the first surface profile from a spatial domain to a frequency domain by a Fast Fourier Transform (FFT) algorithm. Then, a low-pass filter is applied to the first surface profile of the frequency domain to obtain a second surface profile of the frequency domain. Then, the second surface profile is transformed from the frequency domain to the spatial domain by a fast Fourier transform algorithm. Finally, a second surface roughness value is calculated from the second surface profile of the spatial domain.

Referring to fig. 6, fig. 6 shows a first surface roughness value (indicated by diamonds) in different examples, a second surface roughness value (indicated by circles) after performing a laser reflow process with different sets of laser reflow parameter data, and a surface roughness value (experimentally measured) (indicated by squares). In examples 1 to 6, the laser remelting parameter data is a laser energy of 150W and a scanning speed of 600 mm/s; in example 7 and example 8, the laser remelting parameter data is 190W of laser energy and 880mm/s of scanning speed; and in example 9 and example 10, the laser remelting parameter data is a laser energy of 190W and a scanning speed of 700 mm/s. As shown in fig. 6, the surface roughness value is surely reduced after the laser reflow process. Furthermore, the deviation of the difference between the simulated surface roughness value and the measured surface roughness value is not more than 12.3%.

Operation 185 then proceeds to compare the second surface roughness value to a surface roughness threshold. Accordingly, operations 140 through 180 are repeated until the second surface roughness value is less than or equal to the surface roughness threshold. In some embodiments, the surface roughness threshold is 2.8 μm. If the second surface roughness value is greater than the surface roughness threshold, indicating that the set of reflow process parameter data needs to be adjusted, operation 140 is repeated as indicated by the "no" arrow in operation 185. In addition, if the second surface roughness value is less than or equal to the surface roughness threshold value, it indicates that the set of reflow process parameter data is suitable for the reflow process, so operation 190 is performed to perform the laser reflow process on the top surface of the powder layer after the laser powder bed melting process. In some embodiments, the laser reflow process is performed with the set of reflow process parameter data.

Referring to fig. 7, fig. 7 is a flow chart illustrating a method 500 of additive layer manufacturing according to some embodiments of the present disclosure. First, operation 510 is performed to obtain a plurality of first surface roughness values of at least one powder layer, wherein the at least one powder layer is processed by a laser powder bed melting process. In some embodiments, the laser powder bed melting process is performed with a set of melting parameter data, wherein the set of melting parameter data includes a first laser power value, a first scan speed value and a first scan pitch value. In some embodiments, the first surface roughness values are on the same powder layer or on different powder layers. In other embodiments, the first surface roughness values are on different powder layers respectively disposed on different substrates. In some examples, the first surface roughness value is obtained by a virtual metrology method.

Then, an operation 520 is performed to perform a simulation operation with each of the first surface roughness values obtained from the operation 510. Therefore, a reflow operation table containing a plurality of sets of reflow process parameter data can be obtained. The reflow process parameter data are corresponding to the first surface roughness values in a one-to-one manner. In some embodiments, the simulation includes steps similar to operations 120 through 185 described above. For the sake of brevity, the steps of the simulation operation are not described again.

Next, in operation 530, another roughness value of the top surface of the another powder layer is obtained after performing a laser powder bed melting process on the top surface of the another powder layer. In some embodiments, the roughness value is obtained by a virtual metrology method. Then, operation 540 is performed to find another set of reflow process parameter data in the reflow operation table. In other words, another set of reflow process parameter data is determined by using another roughness value. Table 2 is a table of reflow operations according to some embodiments of the present disclosure. For example, if the virtual measurement method yields another roughness value of 2.9 μm, the reflow process parameters should be determined to be 160W of laser energy, a scan speed of 700mm/s and a scan pitch of 100 μm.

TABLE 2

Roughness value (μm)	Laser energy (W)	Scanning speed (mm/s)	Scanning space (mum)
				2.9-4	160	700	100
4-4.5	190	880	100
				>4.5	190	700	100

Then, an operation 550 is performed to perform a laser remelting process on the top surface of the other powder layer after the laser powder bed melting process by using another set of remelting process parameter data. It is understood that if the further roughness value is less than or equal to the roughness threshold (in some embodiments 2.8 μm), the laser remelting process of the further powder layer may be omitted. Again, operations 530 through 550 may be repeated until the desired product is obtained. In other words, the number of times of the laser powder bed melting process and the laser remelting process depends on the final product.

In some embodiments, the methods 100 and 500 may be used to build an Intelligent Additive Manufacturing Architecture (IAMA) to control the surface roughness of each layer of the powder bed during the laser powder bed fusion process. Referring to fig. 8, fig. 8 is a block diagram illustrating an intelligent additive manufacturing architecture 600 according to some embodiments of the present disclosure. The smart additive manufacturing architecture 600 includes an additive manufacturing machine 610, an in-situ metrology system 620, an ex-situ metrology system 630, an automated virtual metrology system 640, an additive manufacturing simulator 650, a smart compensator 660, and a tracking planner 670, wherein the additive manufacturing machine 610 includes a controller 612 and an additive manufacturing tool 614.

FIG. 8 illustrates the basic concept of an intelligent additive manufacturing architecture 600. The in-situ measurement system 620 uses a coaxial high-speed camera and pyrometer to view the molten bath and measure the intensity of radiation emitted by the molten bath, respectively. The external metrology system 630 is configured to collect external data for the production sample, such as melt pool dimensions obtained from the experiment. The tracking planner 670 uses Materialize software to adjust values (PA) based on design requirements, off-board process parameters_f) A recommended range of process Parameters (PR) to produce a planned laser scan path for each layer. The controller 612 of the additive manufacturing machine 610 adjusts the value (PA) according to the on-board process parameter_n) The planned traces are modified and Process Data (PD) having the exact desired energy values and scan velocity values are output to the additive manufacturing facility 614.

A Database (DB)680 collects and manages data generated by the in-situ metrology system 620 (which IS in-situ metrology data IM), data generated by the controller 612 (which IS process parameters PP), data generated by the additive manufacturing tool 614 (which IS in-situ sensing data IS), and data generated by the external metrology system 630 (which IS external metrology data EM), and provides this data to the automated virtual metrology system 640, which then, as described above, the automated virtual metrology system 640 can predict the surface roughness values of the top surface of each layer on the powder bed and transmit to the smart compensator 660, which selects the appropriate laser reflow parameter data from the reflow schedule. If the surface roughness value is greater than the surface roughness threshold (e.g., 2.8 μm), the automated virtual metrology system 640 transfers the surface roughness value to the additive manufacturing simulator 650. Based on the received surface roughness information, the additive manufacturing simulator 659 obtains a desired set of laser remelting parameter data (e.g., laser energy and scan speed) to restore the surface roughness value to less than or equal to the surface roughness threshold. Finally, the set of laser reflow parameter data is transferred to the intelligent compensator 660, which generates control signals to adjust the process parameter settings of the laser reflow process via the tracking planner 670.

In some embodiments, the in-situ measurement system 620 detects manufacturing variations by monitoring the shape change of the molten pool (as observed by a coaxial high-speed CMOS camera) and the reflected light intensity (as measured by a coaxial pyrometer). Further, the laser spot position, puddle image characteristics, and temperature data are transferred to the automated virtual metrology system 640, which then predicts the surface roughness as described above. The in-situ metrology system 620 used in the present disclosure is based on U.S. patent publication No. 20200147893 and U.S. patent publication No. 20200247064. That is, embodiments of the present invention refer to the Incorporated by reference related provisions of these U.S. patent publications.

As described above, the automated virtual metrology system 640 predicts the surface roughness values of each layer in the powder bed based on the data received from the in-situ metrology system 620 and the external metrology system 630. In some embodiments, the automated virtual metrology system 640 uses two predictive algorithms, namely a non-linear Partial Least Squares Regression/Multiple Regression (PLS/MR) algorithm and a linear Neural Network (NN) algorithm. However, other algorithms are also suitable for use in the present disclosure, such as Back Propagation Neural Network (BPNN) algorithm, General Regression Neural Network (GRNN) algorithm, Radial Basis Function Neural Network (RBFNN) algorithm, Simple Regression Network (SRN), Support Vector Data Description (SVDD) algorithm, Support Vector Machine (PLS) algorithm, Multiple Regression (MR) algorithm, Partial Least Squares (Partial Least Squares) algorithm, Nonlinear Least Squares (Nonlinear Least Squares) algorithm, or Nonlinear Linear model (GLLS) algorithm.

Referring to FIG. 9, FIG. 9 is a flow chart illustrating the workflow of an automated virtual metrology system 640 according to some embodiments of the present disclosure. Beginning with operation 710, operation 710 is performed to obtain a set of model data. Then, operation 720 is performed to construct a non-linear Partial Least Squares Regression/Multiple Regression (PLS/MR) model, and operation 740 is performed to construct a linear Neural Network (NN) model. In implementing the linear neural network model, operation 730 is performed before operation 740, wherein operation 740 determines hyper-parameters (i.e., number of epochs (epochs), momentum (momentum), learning rate (learning rate), and number of nodes), which have a large impact on the performance of the neural network. Operation 750 is performed, providing test data. Operation 760 is then performed to predict the surface roughness values using the results of both models. In some embodiments, the hyper-parameter is automatically adjusted using a generalized-genetic algorithm (MGA). The automated virtual metrology system 640 for the present disclosure is based on U.S. patent publication No. 20200147893 and U.S. patent publication No. 20200247064. That is, embodiments of the present invention refer to the Incorporated by reference related provisions of these U.S. patent publications.

According to some embodiments, the laser powder bed fusion process is performed with the intelligent additive manufacturing architecture 600, the method 100, and/or the method 500 such that the average surface roughness is controlled to a level below a surface roughness threshold. Furthermore, the mean and standard deviation of the surface roughness is significantly lower than the samples made by the known laser powder bed fusion process without controlling the surface roughness. Furthermore, in some embodiments, performing the laser powder bed fusion process with the intelligent additive manufacturing architecture 600, the method 100, and/or the method 500 optimizes the average tensile strength of the workpiece from 903MPa to 1013MPa, and reduces the standard deviation from 101.4MPa to 69.5MPa, as compared to the conventional laser powder bed fusion process. Thus, the intelligent additive manufacturing architecture 600, method 100, and/or method 500 can control the quality of the workpiece produced by laser powder bed fusion by controlling the surface roughness of each layer.

The embodiments described above may be implemented using a computer program product that may include a machine-readable medium having stored thereon a plurality of instructions that may be used to program a computer (or other electronic devices) to perform a process according to embodiments of the present disclosure. The machine-readable medium may be, but is not limited to, a floppy disk, an optical disk, a compact-disk-read-only memory (CD-ROM), a magneto-optical disk, a read-only memory (ROM), a Random Access Memory (RAM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), an optical card (optical card) or magnetic card, a flash memory, or any machine-readable medium suitable for storing electronic instructions. Moreover, embodiments of the present invention may also be downloaded as a computer program product, which may be transferred from a remote computer to a requesting computer by way of data signals using a communications connection, such as a network connection.

It should be understood that the foregoing steps or operations in the embodiments of the present disclosure may be combined or omitted, and the order may be adjusted according to actual needs.

Although the present disclosure has been described with reference to several embodiments, it should be understood that the scope of the present disclosure is not limited to the embodiments described above, but rather, may be determined by those skilled in the art.